WIP endpoint abstractions + pruning integration / pruner rewrite

ZFSListChan returns (*DatasetDoesNotExist) for the case mentioned in the comment

.circleci/config.yml

@@ -60,10 +60,10 @@ jobs:
 
       - run: make zrepl-bin
       - run: make vet
-      - run: make test
       - run: make lint
       - run: make release
-
+      - run: make test-go
+      # cannot run test-platform because circle-ci runs in linux containers
 
       - store_artifacts:
           path: ./artifacts/release
@@ -92,11 +92,21 @@ jobs:
             REPO="zrepl/zrepl"
             COMMIT="${CIRCLE_SHA1}"
             JOB_NAME="${CIRCLE_JOB}"
-            curl "https://api.github.com/repos/$REPO/statuses/$COMMIT?access_token=$GITHUB_COMMIT_STATUS_TOKEN" \
+            curl "https://api.github.com/repos/$REPO/statuses/$COMMIT" \
               -H "Content-Type: application/json" \
+              -H "Authorization: token $GITHUB_COMMIT_STATUS_TOKEN" \
               -X POST \
               -d '{"context":"zrepl/publish-ci-artifacts", "state": "success", "description":"CI Build Artifacts for '"$JOB_NAME"'", "target_url":"https://minio.cschwarz.com/minio/zrepl-ci-artifacts/'"$COMMIT"'/"}'
 
+      - run:
+          shell: /bin/bash -euo pipefail
+          command: |
+            # Trigger Debian Package Build
+            curl -v -X POST https://api.github.com/repos/zrepl/debian-binary-packaging/dispatches \
+              -H 'Accept: application/vnd.github.v3+json' \
+              -H "Authorization: token $ZREPL_DEBIAN_BINARYPACKAGIN_TRIGGER_BUILD_GITHUB_TOKEN" \
+              --data '{"event_type": "push", "client_payload": { "zrepl_main_repo_commit": "'"$CIRCLE_SHA1"'", "go_version": "'"${CIRCLE_JOB##build-}"'" }}'
+
   build-1.11:
     <<: *build-latest
     docker:

Makefile

@@ -27,6 +27,8 @@ GO_EXTRA_BUILDFLAGS :=
 GO_BUILDFLAGS := $(GO_MOD_READONLY) $(GO_EXTRA_BUILDFLAGS)
 GO_BUILD := $(GO_ENV_VARS) $(GO) build $(GO_BUILDFLAGS) -ldflags $(GO_LDFLAGS)
 GOLANGCI_LINT := golangci-lint
+GOCOVMERGE := gocovmerge
+
 ifneq ($(GOARM),)
 	ZREPL_TARGET_TUPLE := $(GOOS)-$(GOARCH)v$(GOARM)
 else
@@ -45,7 +47,7 @@ printvars:
 
 release: clean
 	# no cross-platform support for target test
-	$(MAKE) test
+	$(MAKE) test-go
 	$(MAKE) bins-all
 	$(MAKE) noarch
 	$(MAKE) wrapup-and-checksum
@@ -66,6 +68,7 @@ wrapup-and-checksum:
 		-acf $(NOARCH_TARBALL) \
 		$(ARTIFACTDIR)/docs/html \
 		$(ARTIFACTDIR)/bash_completion \
+		$(ARTIFACTDIR)/_zrepl.zsh_completion \
 		$(ARTIFACTDIR)/go_env.txt \
 		dist \
 		config/samples
@@ -98,9 +101,9 @@ clean: docs-clean
 	rm -rf "$(ARTIFACTDIR)"
 
 ##################### BINARIES #####################
-.PHONY: bins-all lint test vet zrepl-bin platformtest-bin
+.PHONY: bins-all lint test-go test-platform cover-merge cover-html vet zrepl-bin test-platform-bin generate-platform-test-list
 
-BINS_ALL_TARGETS := zrepl-bin platformtest-bin vet lint
+BINS_ALL_TARGETS := zrepl-bin test-platform-bin vet lint
 GO_SUPPORTS_ILLUMOS := $(shell $(GO) version | gawk -F '.' '/^go version /{split($$0, comps, " "); split(comps[3], v, "."); if (v[1] == "go1" && v[2] >= 13) { print "illumos"; } else { print "noillumos"; }}')
 bins-all:
 	$(MAKE) $(BINS_ALL_TARGETS) GOOS=freebsd   GOARCH=amd64
@@ -121,27 +124,34 @@ endif
 
 lint:
 	$(GO_ENV_VARS) $(GOLANGCI_LINT) run ./...
-test:
-	$(GO_ENV_VARS) $(GO) test $(GO_BUILDFLAGS) ./...
+
 vet:
 	$(GO_ENV_VARS) $(GO) vet $(GO_BUILDFLAGS) ./...
 
+test-go: $(ARTIFACTDIR)
+	rm -f "$(ARTIFACTDIR)/gotest.cover"
+ifeq ($(COVER),1)
+	$(GO_ENV_VARS) $(GO) test $(GO_BUILDFLAGS) \
+		-coverpkg github.com/zrepl/zrepl/... \
+		-covermode atomic \
+		-coverprofile "$(ARTIFACTDIR)/gotest.cover" \
+		./...
+else
+	$(GO_ENV_VARS) $(GO) test $(GO_BUILDFLAGS) \
+		./...
+endif
+
 zrepl-bin:
 	$(GO_BUILD) -o "$(ARTIFACTDIR)/zrepl-$(ZREPL_TARGET_TUPLE)"
 
-platformtest-bin:
-	$(GO_BUILD) -o "$(ARTIFACTDIR)/platformtest-$(ZREPL_TARGET_TUPLE)" ./platformtest/harness
+generate-platform-test-list:
+	$(GO_BUILD) -o $(ARTIFACTDIR)/generate-platform-test-list ./platformtest/tests/gen
 
-##################### DEV TARGETS #####################
-# not part of the build, must do that manually
-.PHONY: generate format platformtest
-
-generate:
-	protoc -I=replication/logic/pdu --go_out=plugins=grpc:replication/logic/pdu replication/logic/pdu/pdu.proto
-	$(GO_ENV_VARS) $(GO) generate $(GO_BUILDFLAGS) -x ./...
-
-format:
-	goimports -srcdir . -local 'github.com/zrepl/zrepl' -w $(shell find . -type f -name '*.go' -not -path "./vendor/*" -not -name '*.pb.go' -not -name '*_enumer.go')
+test-platform-bin:
+	$(GO_ENV_VARS) $(GO) test $(GO_BUILDFLAGS) \
+		-c -o "$(ARTIFACTDIR)/platformtest-$(ZREPL_TARGET_TUPLE)" \
+		-covermode=atomic -cover -coverpkg github.com/zrepl/zrepl/... \
+		./platformtest/harness
 
 ZREPL_PLATFORMTEST_POOLNAME := zreplplatformtest
 ZREPL_PLATFORMTEST_IMAGEPATH := /tmp/zreplplatformtest.pool.img
@@ -149,18 +159,44 @@ ZREPL_PLATFORMTEST_MOUNTPOINT := /tmp/zreplplatformtest.pool
 ZREPL_PLATFORMTEST_ZFS_LOG := /tmp/zreplplatformtest.zfs.log
 # ZREPL_PLATFORMTEST_STOP_AND_KEEP := -failure.stop-and-keep-pool
 ZREPL_PLATFORMTEST_ARGS := 
-platformtest: # do not track dependency on platformtest-bin to allow build of platformtest outside of test VM
+test-platform: $(ARTIFACTDIR) # do not track dependency on test-platform-bin to allow build of platformtest outside of test VM
 	rm -f "$(ZREPL_PLATFORMTEST_ZFS_LOG)"
+	rm -f "$(ARTIFACTDIR)/platformtest.cover"
 	platformtest/logmockzfs/logzfsenv "$(ZREPL_PLATFORMTEST_ZFS_LOG)" `which zfs` \
 	"$(ARTIFACTDIR)/platformtest-$(ZREPL_TARGET_TUPLE)" \
+		-test.coverprofile "$(ARTIFACTDIR)/platformtest.cover" \
+		-test.v \
+		__DEVEL--i-heard-you-like-tests \
 		-poolname "$(ZREPL_PLATFORMTEST_POOLNAME)" \
 		-imagepath "$(ZREPL_PLATFORMTEST_IMAGEPATH)" \
 		-mountpoint "$(ZREPL_PLATFORMTEST_MOUNTPOINT)" \
 		$(ZREPL_PLATFORMTEST_STOP_AND_KEEP) \
 		$(ZREPL_PLATFORMTEST_ARGS)
 
+cover-merge: $(ARTIFACTDIR)
+	$(GOCOVMERGE) $(ARTIFACTDIR)/platformtest.cover $(ARTIFACTDIR)/gotest.cover > $(ARTIFACTDIR)/merged.cover
+cover-html: cover-merge
+	$(GO) tool cover -html "$(ARTIFACTDIR)/merged.cover" -o "$(ARTIFACTDIR)/merged.cover.html"
+
+test-full:
+	test "$$(id -u)" = "0" || echo "MUST RUN AS ROOT" 1>&2
+	$(MAKE) test-go COVER=1
+	$(MAKE) test-platform
+	$(MAKE) cover-html
+
+##################### DEV TARGETS #####################
+# not part of the build, must do that manually
+.PHONY: generate format
+
+generate: generate-platform-test-list
+	protoc -I=replication/logic/pdu --go_out=plugins=grpc:replication/logic/pdu replication/logic/pdu/pdu.proto
+	$(GO_ENV_VARS) $(GO) generate $(GO_BUILDFLAGS) -x ./...
+
+format:
+	goimports -srcdir . -local 'github.com/zrepl/zrepl' -w $(shell find . -type f -name '*.go' -not -path "./vendor/*" -not -name '*.pb.go' -not -name '*_enumer.go')
+
 ##################### NOARCH #####################
-.PHONY: noarch $(ARTIFACTDIR)/bash_completion $(ARTIFACTDIR)/go_env.txt docs docs-clean
+.PHONY: noarch $(ARTIFACTDIR)/bash_completion $(ARTIFACTDIR)/_zrepl.zsh_completion $(ARTIFACTDIR)/go_env.txt docs docs-clean
 
 
 $(ARTIFACTDIR):
@@ -168,12 +204,16 @@ $(ARTIFACTDIR):
 $(ARTIFACTDIR)/docs: $(ARTIFACTDIR)
 	mkdir -p "$@"
 
-noarch: $(ARTIFACTDIR)/bash_completion $(ARTIFACTDIR)/go_env.txt docs
+noarch: $(ARTIFACTDIR)/bash_completion $(ARTIFACTDIR)/_zrepl.zsh_completion $(ARTIFACTDIR)/go_env.txt docs
 	# pass
 
 $(ARTIFACTDIR)/bash_completion:
 	$(MAKE) zrepl-bin GOOS=$(GOHOSTOS) GOARCH=$(GOHOSTARCH)
-	artifacts/zrepl-$(GOHOSTOS)-$(GOHOSTARCH) bashcomp "$@"
+	artifacts/zrepl-$(GOHOSTOS)-$(GOHOSTARCH) gencompletion bash "$@"
+
+$(ARTIFACTDIR)/_zrepl.zsh_completion:
+	$(MAKE) zrepl-bin GOOS=$(GOHOSTOS) GOARCH=$(GOHOSTARCH)
+	artifacts/zrepl-$(GOHOSTOS)-$(GOHOSTARCH) gencompletion zsh "$@"
 
 $(ARTIFACTDIR)/go_env.txt:
 	$(GO_ENV_VARS) $(GO) env > $@

README.md

@@ -39,7 +39,7 @@ Check out the *Coding Workflow* section below for details.
 * Ship other material provided in `./dist`, e.g. in `/usr/share/zrepl/`.
 * Use `make release ZREPL_VERSION='mydistro-1.2.3_1'`
     * Your distro's name and any versioning supplemental to zrepl's (e.g. package revision) should be in this string
-* Use `make platformtest` **on a test system** to validate that zrepl's abstractions on top of ZFS work with the system ZFS.
+* Use `sudo make test-platform` **on a test system** to validate that zrepl's abstractions on top of ZFS work with the system ZFS.
 * Make sure you are informed about new zrepl versions, e.g. by subscribing to GitHub's release RSS feed.
 
 ## Developer Documentation
@@ -47,10 +47,18 @@ Check out the *Coding Workflow* section below for details.
 zrepl is written in [Go](https://golang.org) and uses [Go modules](https://github.com/golang/go/wiki/Modules) to manage dependencies.
 The documentation is written in [ReStructured Text](http://docutils.sourceforge.net/rst.html) using the [Sphinx](https://www.sphinx-doc.org) framework.
 
-To get started, run `./lazy.sh devsetup` to easily install build dependencies and read `docs/installation.rst -> Compiling from Source`.
+Install **build dependencies** using  `./lazy.sh devsetup`.
 `lazy.sh` uses `python3-pip` to fetch the build dependencies for the docs - you might want to use a [venv](https://docs.python.org/3/library/venv.html).
 If you just want to install the Go dependencies, run `./lazy.sh godep`.
 
+The **test suite** is split into pure **Go tests** (`make test-go`) and **platform tests** that interact with ZFS and thus generally **require root privileges** (`sudo make test-platform`).
+Platform tests run on their own pool with the name `zreplplatformtest`, which is created using the file vdev in `/tmp`.
+
+For a full **code coverage** profile, run `make test-go COVER=1 && sudo make test-platform && make cover-merge`.
+An HTML report can be generated using `make cover-html`.
+
+**Code generation** is triggered by `make generate`. Generated code is committed to the source tree.
+
 ### Project Structure
 
 ```

build/go.mod

@@ -28,6 +28,7 @@ require (
 	github.com/spf13/jwalterweatherman v1.1.0 // indirect
 	github.com/spf13/viper v1.3.2 // indirect
 	github.com/stretchr/testify v1.4.0 // indirect
+	github.com/wadey/gocovmerge v0.0.0-20160331181800-b5bfa59ec0ad // indirect
 	golang.org/x/net v0.0.0-20190613194153-d28f0bde5980 // indirect
 	golang.org/x/sys v0.0.0-20191010194322-b09406accb47 // indirect
 	golang.org/x/tools v0.0.0-20190524210228-3d17549cdc6b

build/go.sum

@@ -214,6 +214,8 @@ github.com/valyala/bytebufferpool v1.0.0/go.mod h1:6bBcMArwyJ5K/AmCkWv1jt77kVWyC
 github.com/valyala/fasthttp v1.2.0/go.mod h1:4vX61m6KN+xDduDNwXrhIAVZaZaZiQ1luJk8LWSxF3s=
 github.com/valyala/quicktemplate v1.1.1/go.mod h1:EH+4AkTd43SvgIbQHYu59/cJyxDoOVRUAfrukLPuGJ4=
 github.com/valyala/tcplisten v0.0.0-20161114210144-ceec8f93295a/go.mod h1:v3UYOV9WzVtRmSR+PDvWpU/qWl4Wa5LApYYX4ZtKbio=
+github.com/wadey/gocovmerge v0.0.0-20160331181800-b5bfa59ec0ad h1:W0LEBv82YCGEtcmPA3uNZBI33/qF//HAAs3MawDjRa0=
+github.com/wadey/gocovmerge v0.0.0-20160331181800-b5bfa59ec0ad/go.mod h1:Hy8o65+MXnS6EwGElrSRjUzQDLXreJlzYLlWiHtt8hM=
 github.com/xordataexchange/crypt v0.0.3-0.20170626215501-b2862e3d0a77/go.mod h1:aYKd//L2LvnjZzWKhF00oedf4jCCReLcmhLdhm1A27Q=
 golang.org/x/crypto v0.0.0-20181203042331-505ab145d0a9/go.mod h1:6SG95UA2DQfeDnfUPMdvaQW0Q7yPrPDi9nlGo2tz2b4=
 golang.org/x/crypto v0.0.0-20190308221718-c2843e01d9a2/go.mod h1:djNgcEr1/C05ACkg1iLfiJU5Ep61QUkGW8qpdssI0+w=

build/tools.go

@@ -6,6 +6,7 @@ import (
 	_ "github.com/alvaroloes/enumer"
 	_ "github.com/golang/protobuf/protoc-gen-go"
 	_ "github.com/golangci/golangci-lint/cmd/golangci-lint"
+	_ "github.com/wadey/gocovmerge"
 	_ "golang.org/x/tools/cmd/goimports"
 	_ "golang.org/x/tools/cmd/stringer"
 )

cli/cli.go

@@ -1,11 +1,13 @@
 package cli
 
 import (
+	"context"
 	"fmt"
 	"os"
 
 	"github.com/spf13/cobra"
 	"github.com/spf13/pflag"
+	"github.com/zrepl/zrepl/daemon/logging/trace"
 
 	"github.com/zrepl/zrepl/config"
 )
@@ -19,29 +21,55 @@ var rootCmd = &cobra.Command{
 	Short: "One-stop ZFS replication solution",
 }
 
-var bashcompCmd = &cobra.Command{
-	Use:   "bashcomp path/to/out/file",
-	Short: "generate bash completions",
-	Run: func(cmd *cobra.Command, args []string) {
-		if len(args) != 1 {
-			fmt.Fprintf(os.Stderr, "specify exactly one positional agument\n")
-			err := cmd.Usage()
-			if err != nil {
-				panic(err)
-			}
-			os.Exit(1)
-		}
-		if err := rootCmd.GenBashCompletionFile(args[0]); err != nil {
-			fmt.Fprintf(os.Stderr, "error generating bash completion: %s", err)
-			os.Exit(1)
-		}
+func init() {
+	rootCmd.PersistentFlags().StringVar(&rootArgs.configPath, "config", "", "config file path")
+}
+
+var genCompletionCmd = &cobra.Command{
+	Use:   "gencompletion",
+	Short: "generate shell auto-completions",
+}
+
+type completionCmdInfo struct {
+	genFunc func(outpath string) error
+	help    string
+}
+
+var completionCmdMap = map[string]completionCmdInfo{
+	"zsh": {
+		rootCmd.GenZshCompletionFile,
+		"  save to file `_zrepl` in your zsh's $fpath",
+	},
+	"bash": {
+		rootCmd.GenBashCompletionFile,
+		"  save to a path and source that path in your .bashrc",
 	},
-	Hidden: true,
 }
 
 func init() {
-	rootCmd.PersistentFlags().StringVar(&rootArgs.configPath, "config", "", "config file path")
-	rootCmd.AddCommand(bashcompCmd)
+	for sh, info := range completionCmdMap {
+		sh, info := sh, info
+		genCompletionCmd.AddCommand(&cobra.Command{
+			Use:     fmt.Sprintf("%s path/to/out/file", sh),
+			Short:   fmt.Sprintf("generate %s completions", sh),
+			Example: info.help,
+			Run: func(cmd *cobra.Command, args []string) {
+				if len(args) != 1 {
+					fmt.Fprintf(os.Stderr, "specify exactly one positional agument\n")
+					err := cmd.Usage()
+					if err != nil {
+						panic(err)
+					}
+					os.Exit(1)
+				}
+				if err := info.genFunc(args[0]); err != nil {
+					fmt.Fprintf(os.Stderr, "error generating %s completion: %s", sh, err)
+					os.Exit(1)
+				}
+			},
+		})
+	}
+	rootCmd.AddCommand(genCompletionCmd)
 }
 
 type Subcommand struct {
@@ -49,7 +77,7 @@ type Subcommand struct {
 	Short            string
 	Example          string
 	NoRequireConfig  bool
-	Run              func(subcommand *Subcommand, args []string) error
+	Run              func(ctx context.Context, subcommand *Subcommand, args []string) error
 	SetupFlags       func(f *pflag.FlagSet)
 	SetupSubcommands func() []*Subcommand
 
@@ -70,7 +98,11 @@ func (s *Subcommand) Config() *config.Config {
 
 func (s *Subcommand) run(cmd *cobra.Command, args []string) {
 	s.tryParseConfig()
-	err := s.Run(s, args)
+	ctx := context.Background()
+	endTask := trace.WithTaskFromStackUpdateCtx(&ctx)
+	defer endTask()
+	err := s.Run(ctx, s, args)
+	endTask()
 	if err != nil {
 		fmt.Fprintf(os.Stderr, "%s\n", err)
 		os.Exit(1)

client/configcheck.go

@@ -1,6 +1,7 @@
 package client
 
 import (
+	"context"
 	"encoding/json"
 	"fmt"
 	"os"
@@ -29,7 +30,7 @@ var ConfigcheckCmd = &cli.Subcommand{
 		f.StringVar(&configcheckArgs.format, "format", "", "dump parsed config object [pretty|yaml|json]")
 		f.StringVar(&configcheckArgs.what, "what", "all", "what to print [all|config|jobs|logging]")
 	},
-	Run: func(subcommand *cli.Subcommand, args []string) error {
+	Run: func(ctx context.Context, subcommand *cli.Subcommand, args []string) error {
 		formatMap := map[string]func(interface{}){
 			"": func(i interface{}) {},
 			"pretty": func(i interface{}) {

client/holds_create.go

@@ -1,14 +0,0 @@
-package client
-
-import "github.com/zrepl/zrepl/cli"
-
-var holdsCmdCreate = &cli.Subcommand{
-	Use:             "create",
-	NoRequireConfig: true,
-	Short:           `create zrepl-managed holds and boomkmarks (for debugging & development only!)`,
-	SetupSubcommands: func() []*cli.Subcommand {
-		return []*cli.Subcommand{
-			holdsCmdCreateStepHold,
-		}
-	},
-}

client/migrate.go

@@ -48,14 +48,13 @@ var migratePlaceholder0_1Args struct {
 	dryRun bool
 }
 
-func doMigratePlaceholder0_1(sc *cli.Subcommand, args []string) error {
+func doMigratePlaceholder0_1(ctx context.Context, sc *cli.Subcommand, args []string) error {
 	if len(args) != 0 {
 		return fmt.Errorf("migration does not take arguments, got %v", args)
 	}
 
 	cfg := sc.Config()
 
-	ctx := context.Background()
 	allFSS, err := zfs.ZFSListMapping(ctx, zfs.NoFilter())
 	if err != nil {
 		return errors.Wrap(err, "cannot list filesystems")
@@ -124,7 +123,7 @@ var fail = color.New(color.FgRed)
 
 var migrateReplicationCursorSkipSentinel = fmt.Errorf("skipping this filesystem")
 
-func doMigrateReplicationCursor(sc *cli.Subcommand, args []string) error {
+func doMigrateReplicationCursor(ctx context.Context, sc *cli.Subcommand, args []string) error {
 	if len(args) != 0 {
 		return fmt.Errorf("migration does not take arguments, got %v", args)
 	}
@@ -137,8 +136,6 @@ func doMigrateReplicationCursor(sc *cli.Subcommand, args []string) error {
 		return fmt.Errorf("exiting migration after error")
 	}
 
-	ctx := context.Background()
-
 	v1cursorJobs := make([]job.Job, 0, len(cfg.Jobs))
 	for i, j := range cfg.Jobs {
 		if jobs[i].Name() != j.Name() {
@@ -211,7 +208,7 @@ func doMigrateReplicationCursorFS(ctx context.Context, v1CursorJobs []job.Job, f
 	}
 	fmt.Printf("identified owning job %q\n", owningJob.Name())
 
-	bookmarks, err := zfs.ZFSListFilesystemVersions(fs, zfs.ListFilesystemVersionsOptions{
+	bookmarks, err := zfs.ZFSListFilesystemVersions(ctx, fs, zfs.ListFilesystemVersionsOptions{
 		Types: zfs.Bookmarks,
 	})
 	if err != nil {

client/pprof.go

@@ -1,67 +1,10 @@
 package client
 
-import (
-	"errors"
-	"log"
-	"os"
-
-	"github.com/zrepl/zrepl/cli"
-	"github.com/zrepl/zrepl/config"
-	"github.com/zrepl/zrepl/daemon"
-)
-
-var pprofArgs struct {
-	daemon.PprofServerControlMsg
-}
+import "github.com/zrepl/zrepl/cli"
 
 var PprofCmd = &cli.Subcommand{
-	Use:   "pprof off | [on TCP_LISTEN_ADDRESS]",
-	Short: "start a http server exposing go-tool-compatible profiling endpoints at TCP_LISTEN_ADDRESS",
-	Run: func(subcommand *cli.Subcommand, args []string) error {
-		if len(args) < 1 {
-			goto enargs
-		}
-		switch args[0] {
-		case "on":
-			pprofArgs.Run = true
-			if len(args) != 2 {
-				return errors.New("must specify TCP_LISTEN_ADDRESS as second positional argument")
-			}
-			pprofArgs.HttpListenAddress = args[1]
-		case "off":
-			if len(args) != 1 {
-				goto enargs
-			}
-			pprofArgs.Run = false
-		}
-
-		RunPProf(subcommand.Config())
-		return nil
-	enargs:
-		return errors.New("invalid number of positional arguments")
-
+	Use: "pprof",
+	SetupSubcommands: func() []*cli.Subcommand {
+		return []*cli.Subcommand{PprofListenCmd, pprofActivityTraceCmd}
 	},
 }
-
-func RunPProf(conf *config.Config) {
-	log := log.New(os.Stderr, "", 0)
-
-	die := func() {
-		log.Printf("exiting after error")
-		os.Exit(1)
-	}
-
-	log.Printf("connecting to zrepl daemon")
-
-	httpc, err := controlHttpClient(conf.Global.Control.SockPath)
-	if err != nil {
-		log.Printf("error creating http client: %s", err)
-		die()
-	}
-	err = jsonRequestResponse(httpc, daemon.ControlJobEndpointPProf, pprofArgs.PprofServerControlMsg, struct{}{})
-	if err != nil {
-		log.Printf("error sending control message: %s", err)
-		die()
-	}
-	log.Printf("finished")
-}

client/pprof_activity_trace.go

@@ -0,0 +1,44 @@
+package client
+
+import (
+	"context"
+	"io"
+	"log"
+	"os"
+
+	"golang.org/x/net/websocket"
+
+	"github.com/zrepl/zrepl/cli"
+)
+
+var pprofActivityTraceCmd = &cli.Subcommand{
+	Use:   "activity-trace ZREPL_PPROF_HOST:ZREPL_PPROF_PORT",
+	Short: "attach to zrepl daemon with activated pprof listener and dump an activity-trace to stdout",
+	Run:   runPProfActivityTrace,
+}
+
+func runPProfActivityTrace(ctx context.Context, subcommand *cli.Subcommand, args []string) error {
+	log := log.New(os.Stderr, "", 0)
+
+	die := func() {
+		log.Printf("exiting after error")
+		os.Exit(1)
+	}
+
+	if len(args) != 1 {
+		log.Printf("exactly one positional argument is required")
+		die()
+	}
+
+	url := "ws://" + args[0] + "/debug/zrepl/activity-trace" // FIXME dont' repeat that
+
+	log.Printf("attaching to activity trace stream %s", url)
+	ws, err := websocket.Dial(url, "", url)
+	if err != nil {
+		log.Printf("error: %s", err)
+		die()
+	}
+
+	_, err = io.Copy(os.Stdout, ws)
+	return err
+}

client/pprof_listen.go

@@ -0,0 +1,68 @@
+package client
+
+import (
+	"context"
+	"errors"
+	"log"
+	"os"
+
+	"github.com/zrepl/zrepl/cli"
+	"github.com/zrepl/zrepl/config"
+	"github.com/zrepl/zrepl/daemon"
+)
+
+var pprofListenCmd struct {
+	daemon.PprofServerControlMsg
+}
+
+var PprofListenCmd = &cli.Subcommand{
+	Use:   "listen off | [on TCP_LISTEN_ADDRESS]",
+	Short: "start a http server exposing go-tool-compatible profiling endpoints at TCP_LISTEN_ADDRESS",
+	Run: func(ctx context.Context, subcommand *cli.Subcommand, args []string) error {
+		if len(args) < 1 {
+			goto enargs
+		}
+		switch args[0] {
+		case "on":
+			pprofListenCmd.Run = true
+			if len(args) != 2 {
+				return errors.New("must specify TCP_LISTEN_ADDRESS as second positional argument")
+			}
+			pprofListenCmd.HttpListenAddress = args[1]
+		case "off":
+			if len(args) != 1 {
+				goto enargs
+			}
+			pprofListenCmd.Run = false
+		}
+
+		RunPProf(subcommand.Config())
+		return nil
+	enargs:
+		return errors.New("invalid number of positional arguments")
+
+	},
+}
+
+func RunPProf(conf *config.Config) {
+	log := log.New(os.Stderr, "", 0)
+
+	die := func() {
+		log.Printf("exiting after error")
+		os.Exit(1)
+	}
+
+	log.Printf("connecting to zrepl daemon")
+
+	httpc, err := controlHttpClient(conf.Global.Control.SockPath)
+	if err != nil {
+		log.Printf("error creating http client: %s", err)
+		die()
+	}
+	err = jsonRequestResponse(httpc, daemon.ControlJobEndpointPProf, pprofListenCmd.PprofServerControlMsg, struct{}{})
+	if err != nil {
+		log.Printf("error sending control message: %s", err)
+		die()
+	}
+	log.Printf("finished")
+}

client/signal.go

@@ -1,6 +1,8 @@
 package client
 
 import (
+	"context"
+
 	"github.com/pkg/errors"
 
 	"github.com/zrepl/zrepl/cli"
@@ -11,7 +13,7 @@ import (
 var SignalCmd = &cli.Subcommand{
 	Use:   "signal [wakeup|reset] JOB",
 	Short: "wake up a job from wait state or abort its current invocation",
-	Run: func(subcommand *cli.Subcommand, args []string) error {
+	Run: func(ctx context.Context, subcommand *cli.Subcommand, args []string) error {
 		return runSignalCmd(subcommand.Config(), args)
 	},
 }

client/status.go

@@ -1,6 +1,7 @@
 package client
 
 import (
+	"context"
 	"fmt"
 	"io"
 	"math"
@@ -180,7 +181,7 @@ var StatusCmd = &cli.Subcommand{
 	Run: runStatus,
 }
 
-func runStatus(s *cli.Subcommand, args []string) error {
+func runStatus(ctx context.Context, s *cli.Subcommand, args []string) error {
 	httpc, err := controlHttpClient(s.Config().Global.Control.SockPath)
 	if err != nil {
 		return err

client/stdinserver.go

@@ -17,7 +17,7 @@ import (
 var StdinserverCmd = &cli.Subcommand{
 	Use:   "stdinserver CLIENT_IDENTITY",
 	Short: "stdinserver transport mode (started from authorized_keys file as forced command)",
-	Run: func(subcommand *cli.Subcommand, args []string) error {
+	Run: func(ctx context.Context, subcommand *cli.Subcommand, args []string) error {
 		return runStdinserver(subcommand.Config(), args)
 	},
 }

client/testcmd.go

@@ -39,7 +39,7 @@ var testFilter = &cli.Subcommand{
 	Run: runTestFilterCmd,
 }
 
-func runTestFilterCmd(subcommand *cli.Subcommand, args []string) error {
+func runTestFilterCmd(ctx context.Context, subcommand *cli.Subcommand, args []string) error {
 
 	if testFilterArgs.job == "" {
 		return fmt.Errorf("must specify --job flag")
@@ -49,7 +49,6 @@ func runTestFilterCmd(subcommand *cli.Subcommand, args []string) error {
 	}
 
 	conf := subcommand.Config()
-	ctx := context.Background()
 
 	var confFilter config.FilesystemsFilter
 	job, err := conf.Job(testFilterArgs.job)
@@ -137,10 +136,9 @@ var testPlaceholder = &cli.Subcommand{
 	Run: runTestPlaceholder,
 }
 
-func runTestPlaceholder(subcommand *cli.Subcommand, args []string) error {
+func runTestPlaceholder(ctx context.Context, subcommand *cli.Subcommand, args []string) error {
 
 	var checkDPs []*zfs.DatasetPath
-	ctx := context.Background()
 
 	// all actions first
 	if testPlaceholderArgs.all {
@@ -197,11 +195,11 @@ var testDecodeResumeToken = &cli.Subcommand{
 	Run: runTestDecodeResumeTokenCmd,
 }
 
-func runTestDecodeResumeTokenCmd(subcommand *cli.Subcommand, args []string) error {
+func runTestDecodeResumeTokenCmd(ctx context.Context, subcommand *cli.Subcommand, args []string) error {
 	if testDecodeResumeTokenArgs.token == "" {
 		return fmt.Errorf("token argument must be specified")
 	}
-	token, err := zfs.ParseResumeToken(context.Background(), testDecodeResumeTokenArgs.token)
+	token, err := zfs.ParseResumeToken(ctx, testDecodeResumeTokenArgs.token)
 	if err != nil {
 		return err
 	}

client/version.go

@@ -1,6 +1,7 @@
 package client
 
 import (
+	"context"
 	"fmt"
 	"os"
 
@@ -25,7 +26,7 @@ var VersionCmd = &cli.Subcommand{
 	SetupFlags: func(f *pflag.FlagSet) {
 		f.StringVar(&versionArgs.Show, "show", "", "version info to show (client|daemon)")
 	},
-	Run: func(subcommand *cli.Subcommand, args []string) error {
+	Run: func(ctx context.Context, subcommand *cli.Subcommand, args []string) error {
 		versionArgs.Config = subcommand.Config()
 		versionArgs.ConfigErr = subcommand.ConfigParsingError()
 		return runVersionCmd()

client/zfsabstractions.go

@@ -14,15 +14,15 @@ import (
 )
 
 var (
-	HoldsCmd = &cli.Subcommand{
-		Use:   "holds",
-		Short: "manage holds & step bookmarks",
+	ZFSAbstractionsCmd = &cli.Subcommand{
+		Use:   "zfs-abstraction",
+		Short: "manage abstractions that zrepl builds on top of ZFS",
 		SetupSubcommands: func() []*cli.Subcommand {
 			return []*cli.Subcommand{
-				holdsCmdList,
-				holdsCmdReleaseAll,
-				holdsCmdReleaseStale,
-				holdsCmdCreate,
+				zabsCmdList,
+				zabsCmdReleaseAll,
+				zabsCmdReleaseStale,
+				zabsCmdCreate,
 			}
 		},
 	}
@@ -30,7 +30,7 @@ var (
 
 // a common set of CLI flags that map to the fields of an
 // endpoint.ListZFSHoldsAndBookmarksQuery
-type holdsFilterFlags struct {
+type zabsFilterFlags struct {
 	Filesystems FilesystemsFilterFlag
 	Job         JobIDFlag
 	Types       AbstractionTypesFlag
@@ -38,7 +38,7 @@ type holdsFilterFlags struct {
 }
 
 // produce a query from the CLI flags
-func (f holdsFilterFlags) Query() (endpoint.ListZFSHoldsAndBookmarksQuery, error) {
+func (f zabsFilterFlags) Query() (endpoint.ListZFSHoldsAndBookmarksQuery, error) {
 	q := endpoint.ListZFSHoldsAndBookmarksQuery{
 		FS:          f.Filesystems.FlagValue(),
 		What:        f.Types.FlagValue(),
@@ -48,7 +48,7 @@ func (f holdsFilterFlags) Query() (endpoint.ListZFSHoldsAndBookmarksQuery, error
 	return q, q.Validate()
 }
 
-func (f *holdsFilterFlags) registerHoldsFilterFlags(s *pflag.FlagSet, verb string) {
+func (f *zabsFilterFlags) registerZabsFilterFlags(s *pflag.FlagSet, verb string) {
 	// Note: the default value is defined in the .FlagValue methods
 	s.Var(&f.Filesystems, "fs", fmt.Sprintf("only %s holds on the specified filesystem [default: all filesystems] [comma-separated list of <dataset-pattern>:<ok|!> pairs]", verb))
 	s.Var(&f.Job, "job", fmt.Sprintf("only %s holds created by the specified job [default: any job]", verb))

client/zfsabstractions_create.go

@@ -0,0 +1,14 @@
+package client
+
+import "github.com/zrepl/zrepl/cli"
+
+var zabsCmdCreate = &cli.Subcommand{
+	Use:             "create",
+	NoRequireConfig: true,
+	Short:           `create zrepl ZFS abstractions (mostly useful for debugging & development, users should not need to use this command)`,
+	SetupSubcommands: func() []*cli.Subcommand {
+		return []*cli.Subcommand{
+			zabsCmdCreateStepHold,
+		}
+	},
+}

client/zfsabstractions_create_step_hold.go

@@ -12,28 +12,28 @@ import (
 	"github.com/zrepl/zrepl/zfs"
 )
 
-var holdsCreateStepHoldFlags struct {
+var zabsCreateStepHoldFlags struct {
 	target string
 	jobid  JobIDFlag
 }
 
-var holdsCmdCreateStepHold = &cli.Subcommand{
+var zabsCmdCreateStepHold = &cli.Subcommand{
 	Use:             "step",
-	Run:             doHoldsCreateStep,
+	Run:             doZabsCreateStep,
 	NoRequireConfig: true,
 	Short:           `create a step hold or bookmark`,
 	SetupFlags: func(f *pflag.FlagSet) {
-		f.StringVarP(&holdsCreateStepHoldFlags.target, "target", "t", "", "snapshot to be held / bookmark to be held")
-		f.VarP(&holdsCreateStepHoldFlags.jobid, "jobid", "j", "jobid for which the hold is installed")
+		f.StringVarP(&zabsCreateStepHoldFlags.target, "target", "t", "", "snapshot to be held / bookmark to be held")
+		f.VarP(&zabsCreateStepHoldFlags.jobid, "jobid", "j", "jobid for which the hold is installed")
 	},
 }
 
-func doHoldsCreateStep(sc *cli.Subcommand, args []string) error {
+func doZabsCreateStep(ctx context.Context, sc *cli.Subcommand, args []string) error {
 	if len(args) > 0 {
 		return errors.New("subcommand takes no arguments")
 	}
 
-	f := &holdsCreateStepHoldFlags
+	f := &zabsCreateStepHoldFlags
 
 	fs, _, _, err := zfs.DecomposeVersionString(f.target)
 	if err != nil {
@@ -44,8 +44,6 @@ func doHoldsCreateStep(sc *cli.Subcommand, args []string) error {
 		return errors.Errorf("jobid must be set")
 	}
 
-	ctx := context.Background()
-
 	v, err := zfs.ZFSGetFilesystemVersion(ctx, f.target)
 	if err != nil {
 		return errors.Wrapf(err, "get info about target %q", f.target)

client/zfsabstractions_list.go

@@ -16,31 +16,30 @@ import (
 	"github.com/zrepl/zrepl/util/chainlock"
 )
 
-var holdsListFlags struct {
-	Filter holdsFilterFlags
+var zabsListFlags struct {
+	Filter zabsFilterFlags
 	Json   bool
 }
 
-var holdsCmdList = &cli.Subcommand{
+var zabsCmdList = &cli.Subcommand{
 	Use:             "list",
-	Run:             doHoldsList,
+	Short:           `list zrepl ZFS abstractions`,
+	Run:             doZabsList,
 	NoRequireConfig: true,
-	Short:           "list holds and bookmarks",
 	SetupFlags: func(f *pflag.FlagSet) {
-		holdsListFlags.Filter.registerHoldsFilterFlags(f, "list")
-		f.BoolVar(&holdsListFlags.Json, "json", false, "emit JSON")
+		zabsListFlags.Filter.registerZabsFilterFlags(f, "list")
+		f.BoolVar(&zabsListFlags.Json, "json", false, "emit JSON")
 	},
 }
 
-func doHoldsList(sc *cli.Subcommand, args []string) error {
+func doZabsList(ctx context.Context, sc *cli.Subcommand, args []string) error {
 	var err error
-	ctx := context.Background()
 
 	if len(args) > 0 {
 		return errors.New("this subcommand takes no positional arguments")
 	}
 
-	q, err := holdsListFlags.Filter.Query()
+	q, err := zabsListFlags.Filter.Query()
 	if err != nil {
 		return errors.Wrap(err, "invalid filter specification on command line")
 	}
@@ -62,7 +61,7 @@ func doHoldsList(sc *cli.Subcommand, args []string) error {
 		for a := range abstractions {
 			func() {
 				defer line.Lock().Unlock()
-				if holdsListFlags.Json {
+				if zabsListFlags.Json {
 					enc.SetIndent("", "  ")
 					if err := enc.Encode(abstractions); err != nil {
 						panic(err)

client/zfsabstractions_release.go

@@ -15,43 +15,42 @@ import (
 )
 
 // shared between release-all and release-step
-var holdsReleaseFlags struct {
-	Filter holdsFilterFlags
+var zabsReleaseFlags struct {
+	Filter zabsFilterFlags
 	Json   bool
 	DryRun bool
 }
 
-func registerHoldsReleaseFlags(s *pflag.FlagSet) {
-	holdsReleaseFlags.Filter.registerHoldsFilterFlags(s, "release")
-	s.BoolVar(&holdsReleaseFlags.Json, "json", false, "emit json instead of pretty-printed")
-	s.BoolVar(&holdsReleaseFlags.DryRun, "dry-run", false, "do a dry-run")
+func registerZabsReleaseFlags(s *pflag.FlagSet) {
+	zabsReleaseFlags.Filter.registerZabsFilterFlags(s, "release")
+	s.BoolVar(&zabsReleaseFlags.Json, "json", false, "emit json instead of pretty-printed")
+	s.BoolVar(&zabsReleaseFlags.DryRun, "dry-run", false, "do a dry-run")
 }
 
-var holdsCmdReleaseAll = &cli.Subcommand{
+var zabsCmdReleaseAll = &cli.Subcommand{
 	Use:             "release-all",
-	Run:             doHoldsReleaseAll,
+	Run:             doZabsReleaseAll,
 	NoRequireConfig: true,
-	Short:           `(DANGEROUS) release all zrepl-managed holds and bookmarks, mostly useful for uninstalling zrepl`,
-	SetupFlags:      registerHoldsReleaseFlags,
+	Short:           `(DANGEROUS) release ALL zrepl ZFS abstractions (mostly useful for uninstalling zrepl completely or for "de-zrepl-ing" a filesystem)`,
+	SetupFlags:      registerZabsReleaseFlags,
 }
 
-var holdsCmdReleaseStale = &cli.Subcommand{
+var zabsCmdReleaseStale = &cli.Subcommand{
 	Use:             "release-stale",
-	Run:             doHoldsReleaseStale,
+	Run:             doZabsReleaseStale,
 	NoRequireConfig: true,
-	Short:           `release stale zrepl-managed holds and boomkarks (useful if zrepl has a bug and doesn't do it by itself)`,
-	SetupFlags:      registerHoldsReleaseFlags,
+	Short:           `release stale zrepl ZFS abstractions (useful if zrepl has a bug and does not do it by itself)`,
+	SetupFlags:      registerZabsReleaseFlags,
 }
 
-func doHoldsReleaseAll(sc *cli.Subcommand, args []string) error {
+func doZabsReleaseAll(ctx context.Context, sc *cli.Subcommand, args []string) error {
 	var err error
-	ctx := context.Background()
 
 	if len(args) > 0 {
 		return errors.New("this subcommand takes no positional arguments")
 	}
 
-	q, err := holdsReleaseFlags.Filter.Query()
+	q, err := zabsReleaseFlags.Filter.Query()
 	if err != nil {
 		return errors.Wrap(err, "invalid filter specification on command line")
 	}
@@ -65,19 +64,18 @@ func doHoldsReleaseAll(sc *cli.Subcommand, args []string) error {
 		// proceed anyways with rest of abstractions
 	}
 
-	return doHoldsRelease_Common(ctx, abstractions)
+	return doZabsRelease_Common(ctx, abstractions)
 }
 
-func doHoldsReleaseStale(sc *cli.Subcommand, args []string) error {
+func doZabsReleaseStale(ctx context.Context, sc *cli.Subcommand, args []string) error {
 
 	var err error
-	ctx := context.Background()
 
 	if len(args) > 0 {
 		return errors.New("this subcommand takes no positional arguments")
 	}
 
-	q, err := holdsReleaseFlags.Filter.Query()
+	q, err := zabsReleaseFlags.Filter.Query()
 	if err != nil {
 		return errors.Wrap(err, "invalid filter specification on command line")
 	}
@@ -87,13 +85,13 @@ func doHoldsReleaseStale(sc *cli.Subcommand, args []string) error {
 		return err // context clear by invocation of command
 	}
 
-	return doHoldsRelease_Common(ctx, stalenessInfo.Stale)
+	return doZabsRelease_Common(ctx, stalenessInfo.Stale)
 }
 
-func doHoldsRelease_Common(ctx context.Context, destroy []endpoint.Abstraction) error {
+func doZabsRelease_Common(ctx context.Context, destroy []endpoint.Abstraction) error {
 
-	if holdsReleaseFlags.DryRun {
-		if holdsReleaseFlags.Json {
+	if zabsReleaseFlags.DryRun {
+		if zabsReleaseFlags.Json {
 			m, err := json.MarshalIndent(destroy, "", "  ")
 			if err != nil {
 				panic(err)
@@ -121,7 +119,7 @@ func doHoldsRelease_Common(ctx context.Context, destroy []endpoint.Abstraction)
 
 	for res := range outcome {
 		hadErr = hadErr || res.DestroyErr != nil
-		if holdsReleaseFlags.Json {
+		if zabsReleaseFlags.Json {
 			err := enc.Encode(res)
 			if err != nil {
 				colorErr.Fprintf(os.Stderr, "cannot marshal there were errors in destroying the abstractions")

config/config.go

@@ -309,6 +309,11 @@ type PruneKeepNotReplicated struct {
 	KeepSnapshotAtCursor bool   `yaml:"keep_snapshot_at_cursor,optional,default=true"`
 }
 
+type PruneKeepStepHolds struct {
+	Type             string   `yaml:"type"`
+	AdditionalJobIds []string `yaml:"additional_job_ids,optional"`
+}
+
 type PruneKeepLastN struct {
 	Type  string `yaml:"type"`
 	Count int    `yaml:"count"`
@@ -480,6 +485,7 @@ func (t *ServeEnum) UnmarshalYAML(u func(interface{}, bool) error) (err error) {
 
 func (t *PruningEnum) UnmarshalYAML(u func(interface{}, bool) error) (err error) {
 	t.Ret, err = enumUnmarshal(u, map[string]interface{}{
+		"step_holds":     &PruneKeepStepHolds{},
 		"not_replicated": &PruneKeepNotReplicated{},
 		"last_n":         &PruneKeepLastN{},
 		"grid":           &PruneGrid{},

config/samples/source.yml

@@ -5,7 +5,11 @@ jobs:
     type: tcp
     listen: "0.0.0.0:8888"
     clients: {
-      "192.168.122.123" : "client1"
+      "192.168.122.123" :               "mysql01",
+      "192.168.122.42" :                "mx01",
+      "2001:0db8:85a3::8a2e:0370:7334": "gateway",
+      "10.23.42.0/24":       "cluster-*",
+      "fde4:8dba:82e1::/64": "san-*",
     }
   filesystems: {
     "<": true,

daemon/daemon.go

@@ -12,6 +12,8 @@ import (
 
 	"github.com/pkg/errors"
 	"github.com/prometheus/client_golang/prometheus"
+	"github.com/zrepl/zrepl/daemon/logging/trace"
+	"github.com/zrepl/zrepl/endpoint"
 
 	"github.com/zrepl/zrepl/config"
 	"github.com/zrepl/zrepl/daemon/job"
@@ -23,9 +25,8 @@ import (
 	"github.com/zrepl/zrepl/zfs/zfscmd"
 )
 
-func Run(conf *config.Config) error {
-
-	ctx, cancel := context.WithCancel(context.Background())
+func Run(ctx context.Context, conf *config.Config) error {
+	ctx, cancel := context.WithCancel(ctx)
 
 	defer cancel()
 	sigChan := make(chan os.Signal, 1)
@@ -39,6 +40,7 @@ func Run(conf *config.Config) error {
 	if err != nil {
 		return errors.Wrap(err, "cannot build logging from config")
 	}
+	outlets.Add(newPrometheusLogOutlet(), logger.Debug)
 
 	confJobs, err := job.JobsFromConfig(conf)
 	if err != nil {
@@ -48,14 +50,23 @@ func Run(conf *config.Config) error {
 	log := logger.NewLogger(outlets, 1*time.Second)
 	log.Info(version.NewZreplVersionInformation().String())
 
+	ctx = logging.WithLoggers(ctx, logging.SubsystemLoggersWithUniversalLogger(log))
+	trace.RegisterCallback(trace.Callback{
+		OnBegin: func(ctx context.Context) { logging.GetLogger(ctx, logging.SubsysTraceData).Debug("begin span") },
+		OnEnd: func(ctx context.Context, spanInfo trace.SpanInfo) {
+			logging.
+				GetLogger(ctx, logging.SubsysTraceData).
+				WithField("duration_s", spanInfo.EndedAt().Sub(spanInfo.StartedAt()).Seconds()).
+				Debug("finished span " + spanInfo.TaskAndSpanStack(trace.SpanStackKindAnnotation))
+		},
+	})
+
 	for _, job := range confJobs {
 		if IsInternalJobName(job.Name()) {
 			panic(fmt.Sprintf("internal job name used for config job '%s'", job.Name())) //FIXME
 		}
 	}
 
-	ctx = job.WithLogger(ctx, log)
-
 	jobs := newJobs()
 
 	// start control socket
@@ -84,6 +95,8 @@ func Run(conf *config.Config) error {
 
 	// register global (=non job-local) metrics
 	zfscmd.RegisterMetrics(prometheus.DefaultRegisterer)
+	trace.RegisterMetrics(prometheus.DefaultRegisterer)
+	endpoint.RegisterMetrics(prometheus.DefaultRegisterer)
 
 	log.Info("starting daemon")
 
@@ -98,6 +111,8 @@ func Run(conf *config.Config) error {
 	case <-ctx.Done():
 		log.WithError(ctx.Err()).Info("context finished")
 	}
+	log.Info("waiting for jobs to finish")
+	<-jobs.wait()
 	log.Info("daemon exiting")
 	return nil
 }
@@ -120,14 +135,11 @@ func newJobs() *jobs {
 	}
 }
 
-const (
-	logJobField string = "job"
-)
-
 func (s *jobs) wait() <-chan struct{} {
 	ch := make(chan struct{})
 	go func() {
 		s.wg.Wait()
+		close(ch)
 	}()
 	return ch
 }
@@ -202,9 +214,8 @@ func (s *jobs) start(ctx context.Context, j job.Job, internal bool) {
 	s.m.Lock()
 	defer s.m.Unlock()
 
-	jobLog := job.GetLogger(ctx).
-		WithField(logJobField, j.Name()).
-		WithOutlet(newPrometheusLogOutlet(j.Name()), logger.Debug)
+	ctx = logging.WithInjectedField(ctx, logging.JobField, j.Name())
+
 	jobName := j.Name()
 	if !internal && IsInternalJobName(jobName) {
 		panic(fmt.Sprintf("internal job name used for non-internal job %s", jobName))
@@ -219,7 +230,6 @@ func (s *jobs) start(ctx context.Context, j job.Job, internal bool) {
 	j.RegisterMetrics(prometheus.DefaultRegisterer)
 
 	s.jobs[jobName] = j
-	ctx = job.WithLogger(ctx, jobLog)
 	ctx = zfscmd.WithJobID(ctx, j.Name())
 	ctx, wakeup := wakeup.Context(ctx)
 	ctx, resetFunc := reset.Context(ctx)
@@ -229,8 +239,8 @@ func (s *jobs) start(ctx context.Context, j job.Job, internal bool) {
 	s.wg.Add(1)
 	go func() {
 		defer s.wg.Done()
-		jobLog.Info("starting job")
-		defer jobLog.Info("job exited")
+		job.GetLogger(ctx).Info("starting job")
+		defer job.GetLogger(ctx).Info("job exited")
 		j.Run(ctx)
 	}()
 }

daemon/hooks/hook_logging.go

@@ -6,29 +6,17 @@ import (
 	"context"
 	"sync"
 
+	"github.com/zrepl/zrepl/daemon/logging"
 	"github.com/zrepl/zrepl/logger"
 	"github.com/zrepl/zrepl/util/envconst"
 )
 
-type contextKey int
-
-const (
-	contextKeyLog contextKey = 0
-)
-
 type Logger = logger.Logger
 
-func WithLogger(ctx context.Context, log Logger) context.Context {
-	return context.WithValue(ctx, contextKeyLog, log)
-}
-
 func GetLogger(ctx context.Context) Logger { return getLogger(ctx) }
 
 func getLogger(ctx context.Context) Logger {
-	if log, ok := ctx.Value(contextKeyLog).(Logger); ok {
-		return log
-	}
-	return logger.NewNullLogger()
+	return logging.GetLogger(ctx, logging.SubsysHooks)
 }
 
 const MAX_HOOK_LOG_SIZE_DEFAULT int = 1 << 20

daemon/hooks/hooks_test.go

@@ -10,9 +10,11 @@ import (
 	"text/template"
 
 	"github.com/stretchr/testify/require"
+	"github.com/zrepl/zrepl/daemon/logging/trace"
 
 	"github.com/zrepl/zrepl/config"
 	"github.com/zrepl/zrepl/daemon/hooks"
+	"github.com/zrepl/zrepl/daemon/logging"
 	"github.com/zrepl/zrepl/logger"
 	"github.com/zrepl/zrepl/zfs"
 )
@@ -69,6 +71,9 @@ func curry(f comparisonAssertionFunc, expected interface{}, right bool) (ret val
 }
 
 func TestHooks(t *testing.T) {
+	ctx, end := trace.WithTaskFromStack(context.Background())
+	defer end()
+
 	testFSName := "testpool/testdataset"
 	testSnapshotName := "testsnap"
 
@@ -418,9 +423,8 @@ jobs:
 
 			cbReached = false
 
-			ctx := context.Background()
 			if testing.Verbose() && !tt.SuppressOutput {
-				ctx = hooks.WithLogger(ctx, log)
+				ctx = logging.WithLoggers(ctx, logging.SubsystemLoggersWithUniversalLogger(log))
 			}
 			plan.Run(ctx, false)
 			report := plan.Report()

daemon/job/active.go

@@ -8,12 +8,13 @@ import (
 
 	"github.com/pkg/errors"
 	"github.com/prometheus/client_golang/prometheus"
+	"github.com/prometheus/common/log"
+	"github.com/zrepl/zrepl/daemon/logging/trace"
 
 	"github.com/zrepl/zrepl/config"
 	"github.com/zrepl/zrepl/daemon/filters"
 	"github.com/zrepl/zrepl/daemon/job/reset"
 	"github.com/zrepl/zrepl/daemon/job/wakeup"
-	"github.com/zrepl/zrepl/daemon/logging"
 	"github.com/zrepl/zrepl/daemon/pruner"
 	"github.com/zrepl/zrepl/daemon/snapper"
 	"github.com/zrepl/zrepl/endpoint"
@@ -79,7 +80,7 @@ func (a *ActiveSide) updateTasks(u func(*activeSideTasks)) activeSideTasks {
 }
 
 type activeMode interface {
-	ConnectEndpoints(rpcLoggers rpc.Loggers, connecter transport.Connecter)
+	ConnectEndpoints(ctx context.Context, connecter transport.Connecter)
 	DisconnectEndpoints()
 	SenderReceiver() (logic.Sender, logic.Receiver)
 	Type() Type
@@ -98,14 +99,14 @@ type modePush struct {
 	snapper       *snapper.PeriodicOrManual
 }
 
-func (m *modePush) ConnectEndpoints(loggers rpc.Loggers, connecter transport.Connecter) {
+func (m *modePush) ConnectEndpoints(ctx context.Context, connecter transport.Connecter) {
 	m.setupMtx.Lock()
 	defer m.setupMtx.Unlock()
 	if m.receiver != nil || m.sender != nil {
 		panic("inconsistent use of ConnectEndpoints and DisconnectEndpoints")
 	}
 	m.sender = endpoint.NewSender(*m.senderConfig)
-	m.receiver = rpc.NewClient(connecter, loggers)
+	m.receiver = rpc.NewClient(connecter, rpc.GetLoggersOrPanic(ctx))
 }
 
 func (m *modePush) DisconnectEndpoints() {
@@ -176,14 +177,14 @@ type modePull struct {
 	interval       config.PositiveDurationOrManual
 }
 
-func (m *modePull) ConnectEndpoints(loggers rpc.Loggers, connecter transport.Connecter) {
+func (m *modePull) ConnectEndpoints(ctx context.Context, connecter transport.Connecter) {
 	m.setupMtx.Lock()
 	defer m.setupMtx.Unlock()
 	if m.receiver != nil || m.sender != nil {
 		panic("inconsistent use of ConnectEndpoints and DisconnectEndpoints")
 	}
 	m.receiver = endpoint.NewReceiver(m.receiverConfig)
-	m.sender = rpc.NewClient(connecter, loggers)
+	m.sender = rpc.NewClient(connecter, rpc.GetLoggersOrPanic(ctx))
 }
 
 func (m *modePull) DisconnectEndpoints() {
@@ -375,16 +376,31 @@ func (j *ActiveSide) SenderConfig() *endpoint.SenderConfig {
 	return push.senderConfig
 }
 
+// The active side of a replication uses one end (sender or receiver)
+// directly by method invocation, without going through a transport that
+// provides a client identity.
+// However, in order to avoid the need to distinguish between direct-method-invocating
+// clients and RPC client, we use an invalid client identity as a sentinel value.
+func FakeActiveSideDirectMethodInvocationClientIdentity(jobId endpoint.JobID) string {
+	return fmt.Sprintf("<local><active><job><client><identity><job=%q>", jobId.String())
+}
+
 func (j *ActiveSide) Run(ctx context.Context) {
+	ctx, endTask := trace.WithTaskAndSpan(ctx, "active-side-job", j.Name())
+	defer endTask()
+
+	ctx = context.WithValue(ctx, endpoint.ClientIdentityKey, FakeActiveSideDirectMethodInvocationClientIdentity(j.name))
+
 	log := GetLogger(ctx)
-	ctx = logging.WithSubsystemLoggers(ctx, log)
 
 	defer log.Info("job exiting")
 
 	periodicDone := make(chan struct{})
 	ctx, cancel := context.WithCancel(ctx)
 	defer cancel()
-	go j.mode.RunPeriodic(ctx, periodicDone)
+	periodicCtx, endTask := trace.WithTask(ctx, "periodic")
+	defer endTask()
+	go j.mode.RunPeriodic(periodicCtx, periodicDone)
 
 	invocationCount := 0
 outer:
@@ -400,17 +416,15 @@ outer:
 		case <-periodicDone:
 		}
 		invocationCount++
-		invLog := log.WithField("invocation", invocationCount)
-		j.do(WithLogger(ctx, invLog))
+		invocationCtx, endSpan := trace.WithSpan(ctx, fmt.Sprintf("invocation-%d", invocationCount))
+		j.do(invocationCtx)
+		endSpan()
 	}
 }
 
 func (j *ActiveSide) do(ctx context.Context) {
 
-	log := GetLogger(ctx)
-	ctx = logging.WithSubsystemLoggers(ctx, log)
-	loggers := rpc.GetLoggersOrPanic(ctx) // filled by WithSubsystemLoggers
-	j.mode.ConnectEndpoints(loggers, j.connecter)
+	j.mode.ConnectEndpoints(ctx, j.connecter)
 	defer j.mode.DisconnectEndpoints()
 
 	// allow cancellation of an invocation (this function)
@@ -433,20 +447,22 @@ func (j *ActiveSide) do(ctx context.Context) {
 			return
 		default:
 		}
+		ctx, endSpan := trace.WithSpan(ctx, "replication")
 		ctx, repCancel := context.WithCancel(ctx)
 		var repWait driver.WaitFunc
 		j.updateTasks(func(tasks *activeSideTasks) {
 			// reset it
 			*tasks = activeSideTasks{}
-			tasks.replicationCancel = repCancel
+			tasks.replicationCancel = func() { repCancel(); endSpan() }
 			tasks.replicationReport, repWait = replication.Do(
 				ctx, logic.NewPlanner(j.promRepStateSecs, j.promBytesReplicated, sender, receiver, j.mode.PlannerPolicy()),
 			)
 			tasks.state = ActiveSideReplicating
 		})
-		log.Info("start replication")
+		GetLogger(ctx).Info("start replication")
 		repWait(true) // wait blocking
 		repCancel()   // always cancel to free up context resources
+		endSpan()
 	}
 
 	{
@@ -455,16 +471,18 @@ func (j *ActiveSide) do(ctx context.Context) {
 			return
 		default:
 		}
+		ctx, endSpan := trace.WithSpan(ctx, "prune_sender")
 		ctx, senderCancel := context.WithCancel(ctx)
 		tasks := j.updateTasks(func(tasks *activeSideTasks) {
 			tasks.prunerSender = j.prunerFactory.BuildSenderPruner(ctx, sender, sender)
-			tasks.prunerSenderCancel = senderCancel
+			tasks.prunerSenderCancel = func() { senderCancel(); endSpan() }
 			tasks.state = ActiveSidePruneSender
 		})
-		log.Info("start pruning sender")
+		GetLogger(ctx).Info("start pruning sender")
 		tasks.prunerSender.Prune()
-		log.Info("finished pruning sender")
+		GetLogger(ctx).Info("finished pruning sender")
 		senderCancel()
+		endSpan()
 	}
 	{
 		select {
@@ -472,16 +490,18 @@ func (j *ActiveSide) do(ctx context.Context) {
 			return
 		default:
 		}
+		ctx, endSpan := trace.WithSpan(ctx, "prune_recever")
 		ctx, receiverCancel := context.WithCancel(ctx)
 		tasks := j.updateTasks(func(tasks *activeSideTasks) {
 			tasks.prunerReceiver = j.prunerFactory.BuildReceiverPruner(ctx, receiver, sender)
-			tasks.prunerReceiverCancel = receiverCancel
+			tasks.prunerReceiverCancel = func() { receiverCancel(); endSpan() }
 			tasks.state = ActiveSidePruneReceiver
 		})
-		log.Info("start pruning receiver")
+		GetLogger(ctx).Info("start pruning receiver")
 		tasks.prunerReceiver.Prune()
-		log.Info("finished pruning receiver")
+		GetLogger(ctx).Info("finished pruning receiver")
 		receiverCancel()
+		endSpan()
 	}
 
 	j.updateTasks(func(tasks *activeSideTasks) {

daemon/job/active_test.go

@@ -0,0 +1,19 @@
+package job
+
+import (
+	"testing"
+
+	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
+	"github.com/zrepl/zrepl/endpoint"
+	"github.com/zrepl/zrepl/transport"
+)
+
+func TestFakeActiveSideDirectMethodInvocationClientIdentityDoesNotPassValidityTest(t *testing.T) {
+	jobid, err := endpoint.MakeJobID("validjobname")
+	require.NoError(t, err)
+	clientIdentity := FakeActiveSideDirectMethodInvocationClientIdentity(jobid)
+	t.Logf("%v", clientIdentity)
+	err = transport.ValidateClientIdentity(clientIdentity)
+	assert.Error(t, err)
+}

daemon/job/job.go

@@ -7,6 +7,7 @@ import (
 
 	"github.com/prometheus/client_golang/prometheus"
 
+	"github.com/zrepl/zrepl/daemon/logging"
 	"github.com/zrepl/zrepl/endpoint"
 	"github.com/zrepl/zrepl/logger"
 	"github.com/zrepl/zrepl/zfs"
@@ -14,21 +15,8 @@ import (
 
 type Logger = logger.Logger
 
-type contextKey int
-
-const (
-	contextKeyLog contextKey = iota
-)
-
 func GetLogger(ctx context.Context) Logger {
-	if l, ok := ctx.Value(contextKeyLog).(Logger); ok {
-		return l
-	}
-	return logger.NewNullLogger()
-}
-
-func WithLogger(ctx context.Context, l Logger) context.Context {
-	return context.WithValue(ctx, contextKeyLog, l)
+	return logging.GetLogger(ctx, logging.SubsysJob)
 }
 
 type Job interface {

daemon/job/passive.go

@@ -6,6 +6,7 @@ import (
 
 	"github.com/pkg/errors"
 	"github.com/prometheus/client_golang/prometheus"
+	"github.com/zrepl/zrepl/daemon/logging/trace"
 
 	"github.com/zrepl/zrepl/config"
 	"github.com/zrepl/zrepl/daemon/filters"
@@ -164,12 +165,14 @@ func (j *PassiveSide) SenderConfig() *endpoint.SenderConfig {
 func (*PassiveSide) RegisterMetrics(registerer prometheus.Registerer) {}
 
 func (j *PassiveSide) Run(ctx context.Context) {
-
+	ctx, endTask := trace.WithTaskAndSpan(ctx, "passive-side-job", j.Name())
+	defer endTask()
 	log := GetLogger(ctx)
 	defer log.Info("job exiting")
-	ctx = logging.WithSubsystemLoggers(ctx, log)
 	{
-		ctx, cancel := context.WithCancel(ctx) // shadowing
+		ctx, endTask := trace.WithTask(ctx, "periodic") // shadowing
+		defer endTask()
+		ctx, cancel := context.WithCancel(ctx)
 		defer cancel()
 		go j.mode.RunPeriodic(ctx)
 	}
@@ -179,8 +182,14 @@ func (j *PassiveSide) Run(ctx context.Context) {
 		panic(fmt.Sprintf("implementation error: j.mode.Handler() returned nil: %#v", j))
 	}
 
-	ctxInterceptor := func(handlerCtx context.Context) context.Context {
-		return logging.WithSubsystemLoggers(handlerCtx, log)
+	ctxInterceptor := func(handlerCtx context.Context, info rpc.HandlerContextInterceptorData, handler func(ctx context.Context)) {
+		// the handlerCtx is clean => need to inherit logging and tracing config from job context
+		handlerCtx = logging.WithInherit(handlerCtx, ctx)
+		handlerCtx = trace.WithInherit(handlerCtx, ctx)
+
+		handlerCtx, endTask := trace.WithTaskAndSpan(handlerCtx, "handler", fmt.Sprintf("job=%q client=%q method=%q", j.Name(), info.ClientIdentity(), info.FullMethod()))
+		defer endTask()
+		handler(handlerCtx)
 	}
 
 	rpcLoggers := rpc.GetLoggersOrPanic(ctx) // WithSubsystemLoggers above

daemon/job/snapjob.go

@@ -2,15 +2,16 @@ package job
 
 import (
 	"context"
+	"fmt"
 	"sort"
 
 	"github.com/pkg/errors"
 	"github.com/prometheus/client_golang/prometheus"
+	"github.com/zrepl/zrepl/daemon/logging/trace"
 
 	"github.com/zrepl/zrepl/config"
 	"github.com/zrepl/zrepl/daemon/filters"
 	"github.com/zrepl/zrepl/daemon/job/wakeup"
-	"github.com/zrepl/zrepl/daemon/logging"
 	"github.com/zrepl/zrepl/daemon/pruner"
 	"github.com/zrepl/zrepl/daemon/snapper"
 	"github.com/zrepl/zrepl/endpoint"
@@ -89,15 +90,18 @@ func (j *SnapJob) OwnedDatasetSubtreeRoot() (rfs *zfs.DatasetPath, ok bool) {
 func (j *SnapJob) SenderConfig() *endpoint.SenderConfig { return nil }
 
 func (j *SnapJob) Run(ctx context.Context) {
+	ctx, endTask := trace.WithTaskAndSpan(ctx, "snap-job", j.Name())
+	defer endTask()
 	log := GetLogger(ctx)
-	ctx = logging.WithSubsystemLoggers(ctx, log)
 
 	defer log.Info("job exiting")
 
 	periodicDone := make(chan struct{})
 	ctx, cancel := context.WithCancel(ctx)
 	defer cancel()
-	go j.snapper.Run(ctx, periodicDone)
+	periodicCtx, endTask := trace.WithTask(ctx, "snapshotting")
+	defer endTask()
+	go j.snapper.Run(periodicCtx, periodicDone)
 
 	invocationCount := 0
 outer:
@@ -112,8 +116,10 @@ outer:
 		case <-periodicDone:
 		}
 		invocationCount++
-		invLog := log.WithField("invocation", invocationCount)
-		j.doPrune(WithLogger(ctx, invLog))
+
+		invocationCtx, endSpan := trace.WithSpan(ctx, fmt.Sprintf("invocation-%d", invocationCount))
+		j.doPrune(invocationCtx)
+		endSpan()
 	}
 }
 
@@ -161,8 +167,9 @@ func (h alwaysUpToDateReplicationCursorHistory) ListFilesystems(ctx context.Cont
 }
 
 func (j *SnapJob) doPrune(ctx context.Context) {
+	ctx, endSpan := trace.WithSpan(ctx, "snap-job-do-prune")
+	defer endSpan()
 	log := GetLogger(ctx)
-	ctx = logging.WithSubsystemLoggers(ctx, log)
 	sender := endpoint.NewSender(endpoint.SenderConfig{
 		JobID: j.name,
 		FSF:   j.fsfilter,

daemon/logging/build_logging.go

@@ -9,20 +9,10 @@ import (
 
 	"github.com/mattn/go-isatty"
 	"github.com/pkg/errors"
-
 	"github.com/zrepl/zrepl/config"
-	"github.com/zrepl/zrepl/daemon/hooks"
-	"github.com/zrepl/zrepl/daemon/pruner"
-	"github.com/zrepl/zrepl/daemon/snapper"
-	"github.com/zrepl/zrepl/endpoint"
+	"github.com/zrepl/zrepl/daemon/logging/trace"
 	"github.com/zrepl/zrepl/logger"
-	"github.com/zrepl/zrepl/replication/driver"
-	"github.com/zrepl/zrepl/replication/logic"
-	"github.com/zrepl/zrepl/rpc"
-	"github.com/zrepl/zrepl/rpc/transportmux"
 	"github.com/zrepl/zrepl/tlsconf"
-	"github.com/zrepl/zrepl/transport"
-	"github.com/zrepl/zrepl/zfs/zfscmd"
 )
 
 func OutletsFromConfig(in config.LoggingOutletEnumList) (*logger.Outlets, error) {
@@ -70,6 +60,8 @@ func OutletsFromConfig(in config.LoggingOutletEnumList) (*logger.Outlets, error)
 type Subsystem string
 
 const (
+	SubsysMeta         Subsystem = "meta"
+	SubsysJob          Subsystem = "job"
 	SubsysReplication  Subsystem = "repl"
 	SubsysEndpoint     Subsystem = "endpoint"
 	SubsysPruning      Subsystem = "pruning"
@@ -81,30 +73,103 @@ const (
 	SubsysRPCControl   Subsystem = "rpc.ctrl"
 	SubsysRPCData      Subsystem = "rpc.data"
 	SubsysZFSCmd       Subsystem = "zfs.cmd"
+	SubsysTraceData    Subsystem = "trace.data"
+	SubsysPlatformtest Subsystem = "platformtest"
 )
 
-func WithSubsystemLoggers(ctx context.Context, log logger.Logger) context.Context {
-	ctx = logic.WithLogger(ctx, log.WithField(SubsysField, SubsysReplication))
-	ctx = driver.WithLogger(ctx, log.WithField(SubsysField, SubsysReplication))
-	ctx = endpoint.WithLogger(ctx, log.WithField(SubsysField, SubsysEndpoint))
-	ctx = pruner.WithLogger(ctx, log.WithField(SubsysField, SubsysPruning))
-	ctx = snapper.WithLogger(ctx, log.WithField(SubsysField, SubsysSnapshot))
-	ctx = hooks.WithLogger(ctx, log.WithField(SubsysField, SubsysHooks))
-	ctx = transport.WithLogger(ctx, log.WithField(SubsysField, SubsysTransport))
-	ctx = transportmux.WithLogger(ctx, log.WithField(SubsysField, SubsysTransportMux))
-	ctx = zfscmd.WithLogger(ctx, log.WithField(SubsysField, SubsysZFSCmd))
-	ctx = rpc.WithLoggers(ctx,
-		rpc.Loggers{
-			General: log.WithField(SubsysField, SubsysRPC),
-			Control: log.WithField(SubsysField, SubsysRPCControl),
-			Data:    log.WithField(SubsysField, SubsysRPCData),
-		},
-	)
-	return ctx
+var AllSubsystems = []Subsystem{
+	SubsysMeta,
+	SubsysJob,
+	SubsysReplication,
+	SubsysEndpoint,
+	SubsysPruning,
+	SubsysSnapshot,
+	SubsysHooks,
+	SubsysTransport,
+	SubsysTransportMux,
+	SubsysRPC,
+	SubsysRPCControl,
+	SubsysRPCData,
+	SubsysZFSCmd,
+	SubsysTraceData,
+	SubsysPlatformtest,
 }
 
-func LogSubsystem(log logger.Logger, subsys Subsystem) logger.Logger {
-	return log.ReplaceField(SubsysField, subsys)
+type injectedField struct {
+	field  string
+	value  interface{}
+	parent *injectedField
+}
+
+func WithInjectedField(ctx context.Context, field string, value interface{}) context.Context {
+	var parent *injectedField
+	parentI := ctx.Value(contextKeyInjectedField)
+	if parentI != nil {
+		parent = parentI.(*injectedField)
+	}
+	// TODO sanity-check `field` now
+	this := &injectedField{field, value, parent}
+	return context.WithValue(ctx, contextKeyInjectedField, this)
+}
+
+func iterInjectedFields(ctx context.Context, cb func(field string, value interface{})) {
+	injI := ctx.Value(contextKeyInjectedField)
+	if injI == nil {
+		return
+	}
+	inj := injI.(*injectedField)
+	for ; inj != nil; inj = inj.parent {
+		cb(inj.field, inj.value)
+	}
+}
+
+type SubsystemLoggers map[Subsystem]logger.Logger
+
+func SubsystemLoggersWithUniversalLogger(l logger.Logger) SubsystemLoggers {
+	loggers := make(SubsystemLoggers)
+	for _, s := range AllSubsystems {
+		loggers[s] = l
+	}
+	return loggers
+}
+
+func WithLoggers(ctx context.Context, loggers SubsystemLoggers) context.Context {
+	return context.WithValue(ctx, contextKeyLoggers, loggers)
+}
+
+func GetLoggers(ctx context.Context) SubsystemLoggers {
+	loggers, ok := ctx.Value(contextKeyLoggers).(SubsystemLoggers)
+	if !ok {
+		return nil
+	}
+	return loggers
+}
+
+func GetLogger(ctx context.Context, subsys Subsystem) logger.Logger {
+	return getLoggerImpl(ctx, subsys, true)
+}
+
+func getLoggerImpl(ctx context.Context, subsys Subsystem, panicIfEnded bool) logger.Logger {
+	loggers, ok := ctx.Value(contextKeyLoggers).(SubsystemLoggers)
+	if !ok || loggers == nil {
+		return logger.NewNullLogger()
+	}
+	l, ok := loggers[subsys]
+	if !ok {
+		return logger.NewNullLogger()
+	}
+
+	l = l.WithField(SubsysField, subsys)
+
+	l = l.WithField(SpanField, trace.GetSpanStackOrDefault(ctx, *trace.StackKindId, "NOSPAN"))
+
+	fields := make(logger.Fields)
+	iterInjectedFields(ctx, func(field string, value interface{}) {
+		fields[field] = value
+	})
+	l = l.WithFields(fields)
+
+	return l
 }
 
 func parseLogFormat(i interface{}) (f EntryFormatter, err error) {

daemon/logging/logging_context.go

@@ -0,0 +1,24 @@
+package logging
+
+import "context"
+
+type contextKey int
+
+const (
+	contextKeyLoggers contextKey = 1 + iota
+	contextKeyInjectedField
+)
+
+var contextKeys = []contextKey{
+	contextKeyLoggers,
+	contextKeyInjectedField,
+}
+
+func WithInherit(ctx, inheritFrom context.Context) context.Context {
+	for _, k := range contextKeys {
+		if v := inheritFrom.Value(k); v != nil {
+			ctx = context.WithValue(ctx, k, v) // no shadow
+		}
+	}
+	return ctx
+}

daemon/logging/logging_formatters.go

@@ -22,6 +22,7 @@ const (
 const (
 	JobField    string = "job"
 	SubsysField string = "subsystem"
+	SpanField   string = "span"
 )
 
 type MetadataFlags int64
@@ -85,7 +86,7 @@ func (f *HumanFormatter) Format(e *logger.Entry) (out []byte, err error) {
 		fmt.Fprintf(&line, "[%s]", col.Sprint(e.Level.Short()))
 	}
 
-	prefixFields := []string{JobField, SubsysField}
+	prefixFields := []string{JobField, SubsysField, SpanField}
 	prefixed := make(map[string]bool, len(prefixFields)+2)
 	for _, field := range prefixFields {
 		val, ok := e.Fields[field]
@@ -174,8 +175,8 @@ func (f *LogfmtFormatter) Format(e *logger.Entry) ([]byte, error) {
 	}
 
 	// at least try and put job and task in front
-	prefixed := make(map[string]bool, 2)
-	prefix := []string{JobField, SubsysField}
+	prefixed := make(map[string]bool, 3)
+	prefix := []string{JobField, SubsysField, SpanField}
 	for _, pf := range prefix {
 		v, ok := e.Fields[pf]
 		if !ok {

daemon/logging/trace/trace.go

@@ -0,0 +1,386 @@
+// package trace provides activity tracing via ctx through Tasks and Spans
+//
+// Basic Concepts
+//
+// Tracing can be used to identify where a piece of code spends its time.
+//
+// The Go standard library provides package runtime/trace which is useful to identify CPU bottlenecks or
+// to understand what happens inside the Go runtime.
+// However, it is not ideal for application level tracing, in particular if those traces should be understandable
+// to tech-savvy users (albeit not developers).
+//
+// This package provides the concept of Tasks and Spans to express what activity is happening within an application:
+//
+//  - Neither task nor span is really tangible but instead contained within the context.Context tree
+//  - Tasks represent concurrent activity (i.e. goroutines).
+//  - Spans represent a semantic stack trace within a task.
+//
+// As a consequence, whenever a context is propagated across goroutine boundary, you need to create a child task:
+//
+//   go func(ctx context.Context) {
+//     ctx, endTask = WithTask(ctx, "what-happens-inside-the-child-task")
+//     defer endTask()
+//     // ...
+//   }(ctx)
+//
+// Within the task, you can open up a hierarchy of spans.
+// In contrast to tasks, which have can multiple concurrently running child tasks,
+// spans must nest and not cross the goroutine boundary.
+//
+//  ctx, endSpan = WithSpan(ctx, "copy-dir")
+//  defer endSpan()
+//  for _, f := range dir.Files() {
+//    func() {
+//      ctx, endSpan := WithSpan(ctx, fmt.Sprintf("copy-file %q", f))
+//      defer endspan()
+//      b, _ := ioutil.ReadFile(f)
+//      _ = ioutil.WriteFile(f + ".copy", b, 0600)
+//    }()
+//  }
+//
+// In combination:
+//  ctx, endTask = WithTask(ctx, "copy-dirs")
+//  defer endTask()
+//  for i := range dirs {
+//    go func(dir string) {
+//      ctx, endTask := WithTask(ctx, "copy-dir")
+//      defer endTask()
+//      for _, f := range filesIn(dir) {
+//        func() {
+//          ctx, endSpan := WithSpan(ctx, fmt.Sprintf("copy-file %q", f))
+//          defer endspan()
+//          b, _ := ioutil.ReadFile(f)
+//          _ = ioutil.WriteFile(f + ".copy", b, 0600)
+//        }()
+//      }
+//    }()
+//  }
+//
+// Note that a span ends at the time you call endSpan - not before and not after that.
+// If you violate the stack-like nesting of spans by forgetting an endSpan() invocation,
+// the out-of-order endSpan() will panic.
+//
+// A similar rule applies to the endTask closure returned by WithTask:
+// If a task has live child tasks at the time you call endTask(), the call will panic.
+//
+// Recovering from endSpan() or endTask() panics will corrupt the trace stack and lead to corrupt tracefile output.
+//
+//
+// Best Practices For Naming Tasks And Spans
+//
+// Tasks should always have string constants as names, and must not contain the `#` character. WHy?
+// First, the visualization by chrome://tracing draws a horizontal bar for each task in the trace.
+// Also, the package appends `#NUM` for each concurrently running instance of a task name.
+// Note that the `#NUM` suffix will be reused if a task has ended, in order to avoid an
+// infinite number of horizontal bars in the visualization.
+//
+//
+// Chrome-compatible Tracefile Support
+//
+// The activity trace generated by usage of WithTask and WithSpan can be rendered to a JSON output file
+// that can be loaded into chrome://tracing .
+// Apart from function GetSpanStackOrDefault, this is the main benefit of this package.
+//
+// First, there is a convenience environment variable 'ZREPL_ACTIVITY_TRACE' that can be set to an output path.
+// From process start onward, a trace is written to that path.
+//
+// More consumers can attach to the activity trace through the ChrometraceClientWebsocketHandler websocket handler.
+//
+// If a write error is encountered with any consumer (including the env-var based one), the consumer is closed and
+// will not receive further trace output.
+package trace
+
+import (
+	"context"
+	"fmt"
+	"strings"
+	"time"
+
+	"github.com/pkg/errors"
+	"github.com/prometheus/client_golang/prometheus"
+	"github.com/zrepl/zrepl/util/chainlock"
+)
+
+var metrics struct {
+	activeTasks prometheus.Gauge
+}
+var taskNamer *uniqueConcurrentTaskNamer = newUniqueTaskNamer()
+
+func init() {
+	metrics.activeTasks = prometheus.NewGauge(prometheus.GaugeOpts{
+		Namespace: "zrepl",
+		Subsystem: "trace",
+		Name:      "active_tasks",
+		Help:      "number of active (tracing-level) tasks in the daemon",
+	})
+}
+
+func RegisterMetrics(r prometheus.Registerer) {
+	r.MustRegister(metrics.activeTasks)
+}
+
+type traceNode struct {
+	id         string
+	annotation string
+	parentTask *traceNode
+
+	mtx chainlock.L
+
+	activeChildTasks int32 // only for task nodes, insignificant for span nodes
+	parentSpan       *traceNode
+	activeChildSpan  *traceNode // nil if task or span doesn't have an active child span
+
+	startedAt time.Time
+	endedAt   time.Time
+}
+
+func (s *traceNode) StartedAt() time.Time { return s.startedAt }
+func (s *traceNode) EndedAt() time.Time   { return s.endedAt }
+
+// Returned from WithTask or WithSpan.
+// Must be called once the task or span ends.
+// See package-level docs for nesting rules.
+// Wrong call order / forgetting to call it will result in panics.
+type DoneFunc func()
+
+var ErrTaskStillHasActiveChildTasks = fmt.Errorf("end task: task still has active child tasks")
+
+// Start a new root task or create a child task of an existing task.
+//
+// This is required when starting a new goroutine and
+// passing an existing task context to it.
+//
+// taskName should be a constantand must not contain '#'
+//
+// The implementation ensures that,
+// if multiple tasks with the same name exist simultaneously,
+// a unique suffix is appended to uniquely identify the task opened with this function.
+func WithTask(ctx context.Context, taskName string) (context.Context, DoneFunc) {
+
+	var parentTask *traceNode
+	nodeI := ctx.Value(contextKeyTraceNode)
+	if nodeI != nil {
+		node := nodeI.(*traceNode)
+		if node.parentSpan != nil {
+			parentTask = node.parentTask
+		} else {
+			parentTask = node
+		}
+	}
+	// find the first ancestor that hasn't ended yet (nil if need be)
+	if parentTask != nil {
+		parentTask.mtx.Lock()
+	}
+	for parentTask != nil && !parentTask.endedAt.IsZero() {
+		thisParent := parentTask
+		parentTask = parentTask.parentTask
+		// lock parent first such that it isn't modified by other callers to thisParent
+		if parentTask != nil {
+			parentTask.mtx.Lock()
+		}
+		thisParent.mtx.Unlock()
+	}
+	// invariant: either parentTask != nil and we hold the lock on parentTask, or parentTask is nil
+
+	taskName, taskNameDone := taskNamer.UniqueConcurrentTaskName(taskName)
+
+	this := &traceNode{
+		id:               genID(),
+		annotation:       taskName,
+		parentTask:       parentTask,
+		activeChildTasks: 0,
+		parentSpan:       nil,
+		activeChildSpan:  nil,
+
+		startedAt: time.Now(),
+		endedAt:   time.Time{},
+	}
+
+	if parentTask != nil {
+		this.parentTask.activeChildTasks++
+		parentTask.mtx.Unlock()
+	}
+
+	ctx = context.WithValue(ctx, contextKeyTraceNode, this)
+
+	chrometraceBeginTask(this)
+
+	metrics.activeTasks.Inc()
+
+	endTaskFunc := func() {
+
+		// only hold locks while manipulating the tree
+		// (trace writer might block too long and unlike spans, tasks are updated concurrently)
+		alreadyEnded := func() (alreadyEnded bool) {
+			if this.parentTask != nil {
+				defer this.parentTask.mtx.Lock().Unlock()
+			}
+			defer this.mtx.Lock().Unlock()
+
+			if this.activeChildTasks != 0 {
+				panic(errors.Wrapf(ErrTaskStillHasActiveChildTasks, "end task: %v active child tasks", this.activeChildTasks))
+			}
+
+			// support idempotent task ends
+			if !this.endedAt.IsZero() {
+				return true
+			}
+			this.endedAt = time.Now()
+
+			if this.parentTask != nil {
+				this.parentTask.activeChildTasks--
+				if this.parentTask.activeChildTasks < 0 {
+					panic("impl error: parent task with negative activeChildTasks count")
+				}
+			}
+			return false
+		}()
+		if alreadyEnded {
+			return
+		}
+
+		chrometraceEndTask(this)
+
+		metrics.activeTasks.Dec()
+
+		taskNameDone()
+	}
+
+	return ctx, endTaskFunc
+}
+
+var ErrAlreadyActiveChildSpan = fmt.Errorf("create child span: span already has an active child span")
+var ErrSpanStillHasActiveChildSpan = fmt.Errorf("end span: span still has active child spans")
+
+// Start a new span.
+// Important: ctx must have an active task (see WithTask)
+func WithSpan(ctx context.Context, annotation string) (context.Context, DoneFunc) {
+	var parentSpan, parentTask *traceNode
+	nodeI := ctx.Value(contextKeyTraceNode)
+	if nodeI != nil {
+		parentSpan = nodeI.(*traceNode)
+		if parentSpan.parentSpan == nil {
+			parentTask = parentSpan
+		} else {
+			parentTask = parentSpan.parentTask
+		}
+	} else {
+		panic("must be called from within a task")
+	}
+
+	this := &traceNode{
+		id:              genID(),
+		annotation:      annotation,
+		parentTask:      parentTask,
+		parentSpan:      parentSpan,
+		activeChildSpan: nil,
+
+		startedAt: time.Now(),
+		endedAt:   time.Time{},
+	}
+
+	parentSpan.mtx.HoldWhile(func() {
+		if parentSpan.activeChildSpan != nil {
+			panic(ErrAlreadyActiveChildSpan)
+		}
+		parentSpan.activeChildSpan = this
+	})
+
+	ctx = context.WithValue(ctx, contextKeyTraceNode, this)
+	chrometraceBeginSpan(this)
+	callbackEndSpan := callbackBeginSpan(ctx)
+
+	endTaskFunc := func() {
+
+		defer parentSpan.mtx.Lock().Unlock()
+		if parentSpan.activeChildSpan != this && this.endedAt.IsZero() {
+			panic("impl error: activeChildSpan should not change while != nil because there can only be one")
+		}
+
+		defer this.mtx.Lock().Unlock()
+		if this.activeChildSpan != nil {
+			panic(ErrSpanStillHasActiveChildSpan)
+		}
+
+		if !this.endedAt.IsZero() {
+			return // support idempotent span ends
+		}
+
+		parentSpan.activeChildSpan = nil
+		this.endedAt = time.Now()
+
+		chrometraceEndSpan(this)
+		callbackEndSpan(this)
+	}
+
+	return ctx, endTaskFunc
+}
+
+type StackKind struct {
+	symbolizeTask func(t *traceNode) string
+	symbolizeSpan func(s *traceNode) string
+}
+
+var (
+	StackKindId = &StackKind{
+		symbolizeTask: func(t *traceNode) string { return t.id },
+		symbolizeSpan: func(s *traceNode) string { return s.id },
+	}
+	SpanStackKindCombined = &StackKind{
+		symbolizeTask: func(t *traceNode) string { return fmt.Sprintf("(%s %q)", t.id, t.annotation) },
+		symbolizeSpan: func(s *traceNode) string { return fmt.Sprintf("(%s %q)", s.id, s.annotation) },
+	}
+	SpanStackKindAnnotation = &StackKind{
+		symbolizeTask: func(t *traceNode) string { return t.annotation },
+		symbolizeSpan: func(s *traceNode) string { return s.annotation },
+	}
+)
+
+func (n *traceNode) task() *traceNode {
+	task := n.parentTask
+	if n.parentSpan == nil {
+		task = n
+	}
+	return task
+}
+
+func (n *traceNode) TaskName() string {
+	task := n.task()
+	return task.annotation
+}
+
+func (this *traceNode) TaskAndSpanStack(kind *StackKind) (spanIdStack string) {
+	task := this.task()
+
+	var spansInTask []*traceNode
+	for s := this; s != nil; s = s.parentSpan {
+		spansInTask = append(spansInTask, s)
+	}
+
+	var tasks []*traceNode
+	for t := task; t != nil; t = t.parentTask {
+		tasks = append(tasks, t)
+	}
+
+	var taskIdsRev []string
+	for i := len(tasks) - 1; i >= 0; i-- {
+		taskIdsRev = append(taskIdsRev, kind.symbolizeTask(tasks[i]))
+	}
+
+	var spanIdsRev []string
+	for i := len(spansInTask) - 1; i >= 0; i-- {
+		spanIdsRev = append(spanIdsRev, kind.symbolizeSpan(spansInTask[i]))
+	}
+
+	taskStack := strings.Join(taskIdsRev, "$")
+	return fmt.Sprintf("%s$%s", taskStack, strings.Join(spanIdsRev, "."))
+}
+
+func GetSpanStackOrDefault(ctx context.Context, kind StackKind, def string) string {
+	if nI := ctx.Value(contextKeyTraceNode); nI != nil {
+		n := nI.(*traceNode)
+		return n.TaskAndSpanStack(StackKindId)
+	} else {
+		return def
+	}
+}

daemon/logging/trace/trace_callback.go

@@ -0,0 +1,54 @@
+package trace
+
+import (
+	"context"
+	"time"
+
+	"github.com/zrepl/zrepl/util/chainlock"
+)
+
+type SpanInfo interface {
+	StartedAt() time.Time
+	EndedAt() time.Time
+	TaskAndSpanStack(kind *StackKind) string
+}
+
+type Callback struct {
+	OnBegin func(ctx context.Context)
+	OnEnd   func(ctx context.Context, spanInfo SpanInfo)
+}
+
+var callbacks struct {
+	mtx chainlock.L
+	cs  []Callback
+}
+
+func RegisterCallback(c Callback) {
+	callbacks.mtx.HoldWhile(func() {
+		callbacks.cs = append(callbacks.cs, c)
+	})
+}
+
+func callbackBeginSpan(ctx context.Context) func(SpanInfo) {
+	// capture the current state of callbacks into a local variable
+	// this is safe because the slice is append-only and immutable
+
+	// (it is important that a callback registered _after_ callbackBeginSpin is called does not get called on OnEnd)
+
+	var cbs []Callback
+	callbacks.mtx.HoldWhile(func() {
+		cbs = callbacks.cs
+	})
+	for _, cb := range cbs {
+		if cb.OnBegin != nil {
+			cb.OnBegin(ctx)
+		}
+	}
+	return func(spanInfo SpanInfo) {
+		for _, cb := range cbs {
+			if cb.OnEnd != nil {
+				cb.OnEnd(ctx, spanInfo)
+			}
+		}
+	}
+}

daemon/logging/trace/trace_chrometrace.go

@@ -0,0 +1,230 @@
+package trace
+
+// The functions in this file are concerned with the generation
+// of trace files based on the information from WithTask and WithSpan.
+//
+// The emitted trace files are open-ended array of JSON objects
+// that follow the Chrome trace file format:
+//   https://docs.google.com/document/d/1CvAClvFfyA5R-PhYUmn5OOQtYMH4h6I0nSsKchNAySU/preview
+//
+// The emitted JSON can be loaded into Chrome's chrome://tracing view.
+//
+// The trace file can be written to a file whose path is specified in an env file,
+// and be written to web sockets established on ChrometraceHttpHandler
+
+import (
+	"bufio"
+	"bytes"
+	"encoding/json"
+	"fmt"
+	"io"
+	"os"
+	"sync"
+	"sync/atomic"
+
+	"golang.org/x/net/websocket"
+
+	"github.com/zrepl/zrepl/util/envconst"
+)
+
+var chrometracePID string
+
+func init() {
+	var err error
+	chrometracePID, err = os.Hostname()
+	if err != nil {
+		panic(err)
+	}
+	chrometracePID = fmt.Sprintf("%q", chrometracePID)
+}
+
+type chrometraceEvent struct {
+	Cat                       string   `json:"cat,omitempty"`
+	Name                      string   `json:"name"`
+	Stack                     []string `json:"stack,omitempty"`
+	Phase                     string   `json:"ph"`
+	TimestampUnixMicroseconds int64    `json:"ts"`
+	DurationMicroseconds      int64    `json:"dur,omitempty"`
+	Pid                       string   `json:"pid"`
+	Tid                       string   `json:"tid"`
+	Id                        string   `json:"id,omitempty"`
+}
+
+func chrometraceBeginSpan(s *traceNode) {
+	taskName := s.TaskName()
+	chrometraceWrite(chrometraceEvent{
+		Name:                      s.annotation,
+		Phase:                     "B",
+		TimestampUnixMicroseconds: s.startedAt.UnixNano() / 1000,
+		Pid:                       chrometracePID,
+		Tid:                       taskName,
+	})
+}
+
+func chrometraceEndSpan(s *traceNode) {
+	taskName := s.TaskName()
+	chrometraceWrite(chrometraceEvent{
+		Name:                      s.annotation,
+		Phase:                     "E",
+		TimestampUnixMicroseconds: s.endedAt.UnixNano() / 1000,
+		Pid:                       chrometracePID,
+		Tid:                       taskName,
+	})
+}
+
+var chrometraceFlowId uint64
+
+func chrometraceBeginTask(s *traceNode) {
+	chrometraceBeginSpan(s)
+
+	if s.parentTask == nil {
+		return
+	}
+	// beginning of a task that has a parent
+	// => use flow events to link parent and child
+
+	flowId := atomic.AddUint64(&chrometraceFlowId, 1)
+	flowIdStr := fmt.Sprintf("%x", flowId)
+
+	parentTask := s.parentTask.TaskName()
+	chrometraceWrite(chrometraceEvent{
+		Cat:                       "task", // seems to be necessary, otherwise the GUI shows some `indexOf` JS error
+		Name:                      "child-task",
+		Phase:                     "s",
+		TimestampUnixMicroseconds: s.startedAt.UnixNano() / 1000, // yes, the child's timestamp (=> from-point of the flow line is at right x-position of parent's bar)
+		Pid:                       chrometracePID,
+		Tid:                       parentTask,
+		Id:                        flowIdStr,
+	})
+
+	childTask := s.TaskName()
+	if parentTask == childTask {
+		panic(parentTask)
+	}
+	chrometraceWrite(chrometraceEvent{
+		Cat:                       "task", // seems to be necessary, otherwise the GUI shows some `indexOf` JS error
+		Name:                      "child-task",
+		Phase:                     "f",
+		TimestampUnixMicroseconds: s.startedAt.UnixNano() / 1000,
+		Pid:                       chrometracePID,
+		Tid:                       childTask,
+		Id:                        flowIdStr,
+	})
+}
+
+func chrometraceEndTask(s *traceNode) {
+	chrometraceEndSpan(s)
+}
+
+type chrometraceConsumerRegistration struct {
+	w io.Writer
+	// errored must have capacity 1, the writer thread will send to it non-blocking, then close it
+	errored chan error
+}
+
+var chrometraceConsumers struct {
+	register  chan chrometraceConsumerRegistration
+	consumers map[chrometraceConsumerRegistration]bool
+	write     chan []byte
+}
+
+func init() {
+	chrometraceConsumers.register = make(chan chrometraceConsumerRegistration)
+	chrometraceConsumers.consumers = make(map[chrometraceConsumerRegistration]bool)
+	chrometraceConsumers.write = make(chan []byte)
+	go func() {
+		kickConsumer := func(c chrometraceConsumerRegistration, err error) {
+			debug("chrometrace kicking consumer %#v after error %v", c, err)
+			select {
+			case c.errored <- err:
+			default:
+			}
+			close(c.errored)
+			delete(chrometraceConsumers.consumers, c)
+		}
+		for {
+			select {
+			case reg := <-chrometraceConsumers.register:
+				debug("registered chrometrace consumer %#v", reg)
+				chrometraceConsumers.consumers[reg] = true
+				n, err := reg.w.Write([]byte("[\n"))
+				if err != nil {
+					kickConsumer(reg, err)
+				} else if n != 2 {
+					kickConsumer(reg, fmt.Errorf("short write: %v", n))
+				}
+				// successfully registered
+
+			case buf := <-chrometraceConsumers.write:
+				debug("chrometrace write request: %s", string(buf))
+				var r bytes.Reader
+				for c := range chrometraceConsumers.consumers {
+					r.Reset(buf)
+					n, err := io.Copy(c.w, &r)
+					debug("chrometrace wrote n=%v bytes to consumer %#v", n, c)
+					if err != nil {
+						kickConsumer(c, err)
+					}
+				}
+			}
+		}
+	}()
+}
+
+func chrometraceWrite(i interface{}) {
+	var buf bytes.Buffer
+	err := json.NewEncoder(&buf).Encode(i)
+	if err != nil {
+		panic(err)
+	}
+	buf.WriteString(",")
+	chrometraceConsumers.write <- buf.Bytes()
+}
+
+func ChrometraceClientWebsocketHandler(conn *websocket.Conn) {
+	defer conn.Close()
+
+	var wg sync.WaitGroup
+	defer wg.Wait()
+	wg.Add(1)
+	go func() {
+		defer wg.Done()
+		r := bufio.NewReader(conn)
+		_, _, _ = r.ReadLine() // ignore errors
+		conn.Close()
+	}()
+
+	errored := make(chan error, 1)
+	chrometraceConsumers.register <- chrometraceConsumerRegistration{
+		w:       conn,
+		errored: errored,
+	}
+
+	wg.Add(1)
+	go func() {
+		defer wg.Done()
+		<-errored
+		conn.Close()
+	}()
+}
+
+var chrometraceFileConsumerPath = envconst.String("ZREPL_ACTIVITY_TRACE", "")
+
+func init() {
+	if chrometraceFileConsumerPath != "" {
+		var err error
+		f, err := os.Create(chrometraceFileConsumerPath)
+		if err != nil {
+			panic(err)
+		}
+		errored := make(chan error, 1)
+		chrometraceConsumers.register <- chrometraceConsumerRegistration{
+			w:       f,
+			errored: errored,
+		}
+		go func() {
+			<-errored
+			f.Close()
+		}()
+	}
+}

daemon/logging/trace/trace_context.go

@@ -0,0 +1,27 @@
+package trace
+
+import "context"
+
+type contextKey int
+
+const (
+	contextKeyTraceNode contextKey = 1 + iota
+)
+
+var contextKeys = []contextKey{
+	contextKeyTraceNode,
+}
+
+// WithInherit inherits the task hierarchy from inheritFrom into ctx.
+// The returned context is a child of ctx, but its task and span are those of inheritFrom.
+//
+// Note that in most use cases, callers most likely want to call WithTask since it will most likely
+// be in some sort of connection handler context.
+func WithInherit(ctx, inheritFrom context.Context) context.Context {
+	for _, k := range contextKeys {
+		if v := inheritFrom.Value(k); v != nil {
+			ctx = context.WithValue(ctx, k, v) // no shadow
+		}
+	}
+	return ctx
+}

daemon/logging/trace/trace_convenience.go

@@ -0,0 +1,74 @@
+package trace
+
+import (
+	"context"
+	"fmt"
+	"runtime"
+	"strings"
+	"sync"
+)
+
+// use like this:
+//
+//	defer WithSpanFromStackUpdateCtx(&existingCtx)()
+//
+//
+func WithSpanFromStackUpdateCtx(ctx *context.Context) DoneFunc {
+	childSpanCtx, end := WithSpan(*ctx, getMyCallerOrPanic())
+	*ctx = childSpanCtx
+	return end
+}
+
+// derive task name from call stack (caller's name)
+func WithTaskFromStack(ctx context.Context) (context.Context, DoneFunc) {
+	return WithTask(ctx, getMyCallerOrPanic())
+}
+
+// derive task name from call stack (caller's name) and update *ctx
+// to point to be the child task ctx
+func WithTaskFromStackUpdateCtx(ctx *context.Context) DoneFunc {
+	child, end := WithTask(*ctx, getMyCallerOrPanic())
+	*ctx = child
+	return end
+}
+
+// create a task and a span within it in one call
+func WithTaskAndSpan(ctx context.Context, task string, span string) (context.Context, DoneFunc) {
+	ctx, endTask := WithTask(ctx, task)
+	ctx, endSpan := WithSpan(ctx, fmt.Sprintf("%s %s", task, span))
+	return ctx, func() {
+		endSpan()
+		endTask()
+	}
+}
+
+// create a span during which several child tasks are spawned using the `add` function
+func WithTaskGroup(ctx context.Context, taskGroup string) (_ context.Context, add func(f func(context.Context)), waitEnd DoneFunc) {
+	var wg sync.WaitGroup
+	ctx, endSpan := WithSpan(ctx, taskGroup)
+	add = func(f func(context.Context)) {
+		wg.Add(1)
+		defer wg.Done()
+		ctx, endTask := WithTask(ctx, taskGroup)
+		defer endTask()
+		f(ctx)
+	}
+	waitEnd = func() {
+		wg.Wait()
+		endSpan()
+	}
+	return ctx, add, waitEnd
+}
+
+func getMyCallerOrPanic() string {
+	pc, _, _, ok := runtime.Caller(2)
+	if !ok {
+		panic("cannot get caller")
+	}
+	details := runtime.FuncForPC(pc)
+	if ok && details != nil {
+		const prefix = "github.com/zrepl/zrepl"
+		return strings.TrimPrefix(strings.TrimPrefix(details.Name(), prefix), "/")
+	}
+	return ""
+}

daemon/logging/trace/trace_convenience_test.go

@@ -0,0 +1,16 @@
+package trace
+
+import (
+	"testing"
+
+	"github.com/stretchr/testify/assert"
+)
+
+func TestGetCallerOrPanic(t *testing.T) {
+	withStackFromCtxMock := func() string {
+		return getMyCallerOrPanic()
+	}
+	ret := withStackFromCtxMock()
+	// zrepl prefix is stripped
+	assert.Equal(t, "daemon/logging/trace.TestGetCallerOrPanic", ret)
+}

daemon/logging/trace/trace_debug.go

@@ -0,0 +1,15 @@
+package trace
+
+import (
+	"fmt"
+	"os"
+)
+
+const debugEnabled = false
+
+func debug(format string, args ...interface{}) {
+	if !debugEnabled {
+		return
+	}
+	fmt.Fprintf(os.Stderr, format+"\n", args...)
+}

daemon/logging/trace/trace_genID.go

@@ -0,0 +1,47 @@
+package trace
+
+import (
+	"encoding/base64"
+	"math/rand"
+	"os"
+	"strings"
+	"time"
+
+	"github.com/zrepl/zrepl/util/envconst"
+)
+
+var genIdPRNG = rand.New(rand.NewSource(1))
+
+func init() {
+	genIdPRNG.Seed(time.Now().UnixNano())
+	genIdPRNG.Seed(int64(os.Getpid()))
+}
+
+var genIdNumBytes = envconst.Int("ZREPL_TRACE_ID_NUM_BYTES", 3)
+
+func init() {
+	if genIdNumBytes < 1 {
+		panic("trace node id byte length must be at least 1")
+	}
+}
+
+func genID() string {
+	var out strings.Builder
+	enc := base64.NewEncoder(base64.RawStdEncoding, &out)
+	buf := make([]byte, genIdNumBytes)
+	for i := 0; i < len(buf); {
+		n, err := genIdPRNG.Read(buf[i:])
+		if err != nil {
+			panic(err)
+		}
+		i += n
+	}
+	n, err := enc.Write(buf[:])
+	if err != nil || n != len(buf) {
+		panic(err)
+	}
+	if err := enc.Close(); err != nil {
+		panic(err)
+	}
+	return out.String()
+}

daemon/logging/trace/trace_test.go

@@ -0,0 +1,205 @@
+package trace
+
+import (
+	"context"
+	"fmt"
+	"testing"
+
+	"github.com/gitchander/permutation"
+	"github.com/pkg/errors"
+	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
+)
+
+func TestRegularSpanUsage(t *testing.T) {
+	root, endRoot := WithTask(context.Background(), "root")
+	defer endRoot()
+
+	s1, endS1 := WithSpan(root, "parent")
+	s2, endS2 := WithSpan(s1, "child")
+	_, endS3 := WithSpan(s2, "grand-child")
+	require.NotPanics(t, func() { endS3() })
+	require.NotPanics(t, func() { endS2() })
+
+	// reuse
+	_, endS4 := WithSpan(s1, "child-2")
+	require.NotPanics(t, func() { endS4() })
+
+	// close parent
+	require.NotPanics(t, func() { endS1() })
+}
+
+func TestMultipleActiveChildSpansNotAllowed(t *testing.T) {
+	root, endRoot := WithTask(context.Background(), "root")
+	defer endRoot()
+
+	s1, _ := WithSpan(root, "s1")
+	_, endS2 := WithSpan(s1, "s1-child1")
+
+	require.PanicsWithValue(t, ErrAlreadyActiveChildSpan, func() {
+		_, _ = WithSpan(s1, "s1-child2")
+	})
+
+	endS2()
+
+	require.NotPanics(t, func() {
+		_, _ = WithSpan(s1, "s1-child2")
+	})
+}
+
+func TestForkingChildSpansNotAllowed(t *testing.T) {
+	root, endRoot := WithTask(context.Background(), "root")
+	defer endRoot()
+
+	s1, _ := WithSpan(root, "s1")
+	sc, endSC := WithSpan(s1, "s1-child")
+	_, _ = WithSpan(sc, "s1-child-child")
+
+	require.PanicsWithValue(t, ErrSpanStillHasActiveChildSpan, func() {
+		endSC()
+	})
+}
+
+func TestRegularTaskUsage(t *testing.T) {
+	// assert concurrent activities on different tasks can end in any order
+	closeOrder := []int{0, 1, 2}
+	closeOrders := permutation.New(permutation.IntSlice(closeOrder))
+	for closeOrders.Next() {
+		t.Run(fmt.Sprintf("%v", closeOrder), func(t *testing.T) {
+			root, endRoot := WithTask(context.Background(), "root")
+			defer endRoot()
+
+			c1, endC1 := WithTask(root, "c1")
+			defer endC1()
+			c2, endC2 := WithTask(root, "c2")
+			defer endC2()
+
+			// begin 3 concurrent activities
+			_, endAR := WithSpan(root, "aR")
+			_, endAC1 := WithSpan(c1, "aC1")
+			_, endAC2 := WithSpan(c2, "aC2")
+
+			endFuncs := []DoneFunc{endAR, endAC1, endAC2}
+			for _, i := range closeOrder {
+				require.NotPanics(t, func() {
+					endFuncs[i]()
+				}, "%v", i)
+			}
+		})
+	}
+}
+
+func TestTaskEndWithActiveChildTaskNotAllowed(t *testing.T) {
+	root, _ := WithTask(context.Background(), "root")
+	c, endC := WithTask(root, "child")
+	_, _ = WithTask(c, "grand-child")
+	func() {
+		defer func() {
+			r := recover()
+			require.NotNil(t, r)
+			err, ok := r.(error)
+			require.True(t, ok)
+			require.Equal(t, ErrTaskStillHasActiveChildTasks, errors.Cause(err))
+		}()
+		endC()
+	}()
+
+}
+
+func TestIdempotentEndTask(t *testing.T) {
+	_, end := WithTask(context.Background(), "root")
+	end()
+	require.NotPanics(t, func() { end() })
+}
+
+func TestSpansPanicIfNoParentTask(t *testing.T) {
+	require.Panics(t, func() { WithSpan(context.Background(), "taskless-span") })
+}
+
+func TestIdempotentEndSpan(t *testing.T) {
+	root, _ := WithTask(context.Background(), "root")
+	_, end := WithSpan(root, "span")
+	end()
+	require.NotPanics(t, func() { end() })
+}
+
+func logAndGetTraceNode(t *testing.T, descr string, ctx context.Context) *traceNode {
+	n, ok := ctx.Value(contextKeyTraceNode).(*traceNode)
+	require.True(t, ok)
+	t.Logf("% 20s %p %#v", descr, n, n)
+	return n
+}
+
+func TestWhiteboxHierachy(t *testing.T) {
+	root, e1 := WithTask(context.Background(), "root")
+	rootN := logAndGetTraceNode(t, "root", root)
+	assert.Nil(t, rootN.parentTask)
+	assert.Nil(t, rootN.parentSpan)
+
+	child, e2 := WithSpan(root, "child")
+	childN := logAndGetTraceNode(t, "child", child)
+	assert.Equal(t, rootN, childN.parentTask)
+	assert.Equal(t, rootN, childN.parentSpan)
+
+	grandchild, e3 := WithSpan(child, "grandchild")
+	grandchildN := logAndGetTraceNode(t, "grandchild", grandchild)
+	assert.Equal(t, rootN, grandchildN.parentTask)
+	assert.Equal(t, childN, grandchildN.parentSpan)
+
+	gcTask, e4 := WithTask(grandchild, "grandchild-task")
+	gcTaskN := logAndGetTraceNode(t, "grandchild-task", gcTask)
+	assert.Equal(t, rootN, gcTaskN.parentTask)
+	assert.Nil(t, gcTaskN.parentSpan)
+
+	// it is allowed that a child task outlives the _span_ in which it was created
+	// (albeit not its parent task)
+	e3()
+	e2()
+	gcTaskSpan, e5 := WithSpan(gcTask, "granschild-task-span")
+	gcTaskSpanN := logAndGetTraceNode(t, "granschild-task-span", gcTaskSpan)
+	assert.Equal(t, gcTaskN, gcTaskSpanN.parentTask)
+	assert.Equal(t, gcTaskN, gcTaskSpanN.parentSpan)
+	e5()
+
+	e4()
+	e1()
+}
+
+func TestOrphanTasksBecomeNewRootWhenAllAncestorsAreDead(t *testing.T) {
+	parent, e1 := WithTask(context.Background(), "parent")
+	_ = logAndGetTraceNode(t, "parent-task", parent)
+	e1()
+
+	var child context.Context
+	var e2 DoneFunc
+	require.NotPanics(t, func() {
+		child, e2 = WithTask(parent, "child")
+	})
+	childN := logAndGetTraceNode(t, "child-task", child)
+	assert.Nil(t, childN.parentTask)
+
+	grandchild, _ := WithTask(child, "grandchild")
+	grandchildN := logAndGetTraceNode(t, "grandchild-task", grandchild)
+	assert.Equal(t, childN, grandchildN.parentTask)
+
+	require.Panics(t, func() { e2() }, "if the parent was alive at creation of child task, wrong termination order remains a panicable offense though")
+}
+
+func TestOrphanTaskBecomesChildToNearestLiveAncestor(t *testing.T) {
+	parent, e1 := WithTask(context.Background(), "parent")
+	parentN := logAndGetTraceNode(t, "parent-task", parent)
+
+	child, e2 := WithTask(parent, "child")
+	childN := logAndGetTraceNode(t, "child-task", child)
+	assert.Equal(t, parentN, childN.parentTask)
+	e2()
+
+	grandchild, e3 := WithTask(child, "grandchild")
+	grandchildN := logAndGetTraceNode(t, "grandchild-task", grandchild)
+	assert.Equal(t, parentN, grandchildN.parentTask) // child is already dead
+
+	require.NotPanics(t, func() {
+		e3()
+		e1()
+	})
+}

daemon/logging/trace/trace_unique_concurrent_task_namer.go

@@ -0,0 +1,53 @@
+package trace
+
+import (
+	"fmt"
+	"strings"
+	"sync"
+
+	"github.com/willf/bitset"
+)
+
+type uniqueConcurrentTaskNamer struct {
+	mtx    sync.Mutex
+	active map[string]*bitset.BitSet
+}
+
+// bitvecLengthGauge may be nil
+func newUniqueTaskNamer() *uniqueConcurrentTaskNamer {
+	return &uniqueConcurrentTaskNamer{
+		active: make(map[string]*bitset.BitSet),
+	}
+}
+
+// appends `#%d` to `name` such that until `done` is called,
+// it is guaranteed that `#%d` is not returned a second time for the same `name`
+func (namer *uniqueConcurrentTaskNamer) UniqueConcurrentTaskName(name string) (uniqueName string, done func()) {
+	if strings.Contains(name, "#") {
+		panic(name)
+	}
+	namer.mtx.Lock()
+	act, ok := namer.active[name]
+	if !ok {
+		act = bitset.New(64) // FIXME magic const
+		namer.active[name] = act
+	}
+	id, ok := act.NextClear(0)
+	if !ok {
+		// if !ok, all bits are 1 and act.Len() returns the next bit
+		id = act.Len()
+		// FIXME unbounded growth without reclamation
+	}
+	act.Set(id)
+	namer.mtx.Unlock()
+
+	return fmt.Sprintf("%s#%d", name, id), func() {
+		namer.mtx.Lock()
+		defer namer.mtx.Unlock()
+		act, ok := namer.active[name]
+		if !ok {
+			panic("must be initialized upon entry")
+		}
+		act.Clear(id)
+	}
+}

daemon/logging/trace/trace_unique_concurrent_task_namer_test.go

@@ -0,0 +1,58 @@
+package trace
+
+import (
+	"fmt"
+	"sync"
+	"sync/atomic"
+	"testing"
+
+	"github.com/stretchr/testify/require"
+	"github.com/willf/bitset"
+)
+
+func TestBitsetFeaturesForUniqueConcurrentTaskNamer(t *testing.T) {
+	var b bitset.BitSet
+	require.Equal(t, uint(0), b.Len())
+	require.Equal(t, uint(0), b.Count())
+
+	b.Set(0)
+	require.Equal(t, uint(1), b.Len())
+	require.Equal(t, uint(1), b.Count())
+
+	b.Set(8)
+	require.Equal(t, uint(9), b.Len())
+	require.Equal(t, uint(2), b.Count())
+
+	b.Set(1)
+	require.Equal(t, uint(9), b.Len())
+	require.Equal(t, uint(3), b.Count())
+}
+
+func TestUniqueConcurrentTaskNamer(t *testing.T) {
+	namer := newUniqueTaskNamer()
+
+	var wg sync.WaitGroup
+
+	const N = 8128
+	const Q = 23
+
+	var fails uint32
+	var m sync.Map
+	wg.Add(N)
+	for i := 0; i < N; i++ {
+		go func(i int) {
+			defer wg.Done()
+			name := fmt.Sprintf("%d", i/Q)
+			uniqueName, done := namer.UniqueConcurrentTaskName(name)
+			act, _ := m.LoadOrStore(uniqueName, i)
+			if act.(int) != i {
+				atomic.AddUint32(&fails, 1)
+			}
+			m.Delete(uniqueName)
+			done()
+		}(i)
+	}
+	wg.Wait()
+
+	require.Equal(t, uint32(0), fails)
+}

daemon/main.go

@@ -1,6 +1,8 @@
 package daemon
 
 import (
+	"context"
+
 	"github.com/zrepl/zrepl/cli"
 	"github.com/zrepl/zrepl/logger"
 )
@@ -10,7 +12,7 @@ type Logger = logger.Logger
 var DaemonCmd = &cli.Subcommand{
 	Use:   "daemon",
 	Short: "run the zrepl daemon",
-	Run: func(subcommand *cli.Subcommand, args []string) error {
-		return Run(subcommand.Config())
+	Run: func(ctx context.Context, subcommand *cli.Subcommand, args []string) error {
+		return Run(ctx, subcommand.Config())
 	},
 }

daemon/pprof.go

@@ -8,7 +8,11 @@ import (
 	"net"
 	"net/http/pprof"
 
+	"github.com/prometheus/client_golang/prometheus/promhttp"
+	"golang.org/x/net/websocket"
+
 	"github.com/zrepl/zrepl/daemon/job"
+	"github.com/zrepl/zrepl/daemon/logging/trace"
 )
 
 type pprofServer struct {
@@ -64,6 +68,8 @@ outer:
 			mux.Handle("/debug/pprof/profile", http.HandlerFunc(pprof.Profile))
 			mux.Handle("/debug/pprof/symbol", http.HandlerFunc(pprof.Symbol))
 			mux.Handle("/debug/pprof/trace", http.HandlerFunc(pprof.Trace))
+			mux.Handle("/metrics", promhttp.Handler())
+			mux.Handle("/debug/zrepl/activity-trace", websocket.Handler(trace.ChrometraceClientWebsocketHandler))
 			go func() {
 				err := http.Serve(s.listener, mux)
 				if ctx.Err() != nil {

daemon/prometheus.go

@@ -10,6 +10,7 @@ import (
 
 	"github.com/zrepl/zrepl/config"
 	"github.com/zrepl/zrepl/daemon/job"
+	"github.com/zrepl/zrepl/daemon/logging"
 	"github.com/zrepl/zrepl/endpoint"
 	"github.com/zrepl/zrepl/logger"
 	"github.com/zrepl/zrepl/rpc/dataconn/frameconn"
@@ -86,16 +87,19 @@ func (j *prometheusJob) Run(ctx context.Context) {
 }
 
 type prometheusJobOutlet struct {
-	jobName string
 }
 
 var _ logger.Outlet = prometheusJobOutlet{}
 
-func newPrometheusLogOutlet(jobName string) prometheusJobOutlet {
-	return prometheusJobOutlet{jobName}
+func newPrometheusLogOutlet() prometheusJobOutlet {
+	return prometheusJobOutlet{}
 }
 
 func (o prometheusJobOutlet) WriteEntry(entry logger.Entry) error {
-	prom.taskLogEntries.WithLabelValues(o.jobName, entry.Level.String()).Inc()
+	jobFieldVal, ok := entry.Fields[logging.JobField].(string)
+	if !ok {
+		jobFieldVal = "_nojobid"
+	}
+	prom.taskLogEntries.WithLabelValues(jobFieldVal, entry.Level.String()).Inc()
 	return nil
 }

daemon/pruner.v2/fsstate_enumer.go

@@ -0,0 +1,72 @@
+// Code generated by "enumer -type=FSState -json"; DO NOT EDIT.
+
+//
+package pruner
+
+import (
+	"encoding/json"
+	"fmt"
+)
+
+const _FSStateName = "FSStateInitializedFSStatePlanningFSStatePlanErrFSStateExecutingFSStateExecuteErrFSStateExecuteSuccess"
+
+var _FSStateIndex = [...]uint8{0, 18, 33, 47, 63, 80, 101}
+
+func (i FSState) String() string {
+	if i < 0 || i >= FSState(len(_FSStateIndex)-1) {
+		return fmt.Sprintf("FSState(%d)", i)
+	}
+	return _FSStateName[_FSStateIndex[i]:_FSStateIndex[i+1]]
+}
+
+var _FSStateValues = []FSState{0, 1, 2, 3, 4, 5}
+
+var _FSStateNameToValueMap = map[string]FSState{
+	_FSStateName[0:18]:   0,
+	_FSStateName[18:33]:  1,
+	_FSStateName[33:47]:  2,
+	_FSStateName[47:63]:  3,
+	_FSStateName[63:80]:  4,
+	_FSStateName[80:101]: 5,
+}
+
+// FSStateString retrieves an enum value from the enum constants string name.
+// Throws an error if the param is not part of the enum.
+func FSStateString(s string) (FSState, error) {
+	if val, ok := _FSStateNameToValueMap[s]; ok {
+		return val, nil
+	}
+	return 0, fmt.Errorf("%s does not belong to FSState values", s)
+}
+
+// FSStateValues returns all values of the enum
+func FSStateValues() []FSState {
+	return _FSStateValues
+}
+
+// IsAFSState returns "true" if the value is listed in the enum definition. "false" otherwise
+func (i FSState) IsAFSState() bool {
+	for _, v := range _FSStateValues {
+		if i == v {
+			return true
+		}
+	}
+	return false
+}
+
+// MarshalJSON implements the json.Marshaler interface for FSState
+func (i FSState) MarshalJSON() ([]byte, error) {
+	return json.Marshal(i.String())
+}
+
+// UnmarshalJSON implements the json.Unmarshaler interface for FSState
+func (i *FSState) UnmarshalJSON(data []byte) error {
+	var s string
+	if err := json.Unmarshal(data, &s); err != nil {
+		return fmt.Errorf("FSState should be a string, got %s", data)
+	}
+
+	var err error
+	*i, err = FSStateString(s)
+	return err
+}

daemon/pruner.v2/pruner.go

@@ -0,0 +1,454 @@
+package pruner
+
+import (
+	"context"
+	"encoding/json"
+	"fmt"
+	"os"
+
+	"github.com/pkg/errors"
+	"github.com/zrepl/zrepl/daemon/logging"
+	"github.com/zrepl/zrepl/daemon/logging/trace"
+	"github.com/zrepl/zrepl/endpoint"
+	"github.com/zrepl/zrepl/pruning"
+	"github.com/zrepl/zrepl/zfs"
+)
+
+type Pruner struct {
+	fsfilter  endpoint.FSFilter
+	jid       endpoint.JobID
+	side      Side
+	keepRules []pruning.KeepRule
+
+	// all channels consumed by the run loop
+	reportReqs chan reportRequest
+	stopReqs   chan stopRequest
+	done       chan struct{}
+	fsListRes  chan fsListRes
+
+	state State
+
+	listFilesystemsError error       // only in state StateListFilesystemsError
+	fsPruners            []*FSPruner // only in state StateFanOutFilesystems
+}
+
+//go:generate enumer -type=State -json
+type State int
+
+const (
+	StateInitialized State = iota
+	StateListFilesystems
+	StateListFilesystemsError
+	StateFanOutFilesystems
+	StateDone
+)
+
+type Report struct {
+	State                State
+	ListFilesystemsError error       // only valid in StateListFilesystemsError
+	Filesystems          []*FSReport // valid from StateFanOutFilesystems
+}
+
+type reportRequest struct {
+	ctx   context.Context
+	reply chan *Report
+}
+
+type runRequest struct {
+	complete chan struct{}
+}
+
+type stopRequest struct {
+	complete chan struct{}
+}
+
+type fsListRes struct {
+	filesystems []*zfs.DatasetPath
+	err         error
+}
+
+type Side interface {
+	// may return both nil, indicating there is no replication position
+	GetReplicationPosition(ctx context.Context, fs string) (*zfs.FilesystemVersion, error)
+	isSide() Side
+}
+
+func NewPruner(fsfilter endpoint.FSFilter, jid endpoint.JobID, side Side, keepRules []pruning.KeepRule) *Pruner {
+	return &Pruner{
+		fsfilter,
+		jid,
+		side,
+		keepRules,
+		make(chan reportRequest),
+		make(chan stopRequest),
+		make(chan struct{}),
+		make(chan fsListRes),
+		StateInitialized,
+		nil,
+		nil,
+	}
+}
+
+func (p *Pruner) Run(ctx context.Context) *Report {
+
+	ctx, cancel := context.WithCancel(ctx)
+	defer cancel()
+
+	if p.state != StateInitialized {
+		panic("Run can onl[y be called once")
+	}
+
+	go func() {
+		fss, err := zfs.ZFSListMapping(ctx, p.fsfilter)
+		p.fsListRes <- fsListRes{fss, err}
+	}()
+
+	for {
+		select {
+		case res := <-p.fsListRes:
+			if res.err != nil {
+				p.state = StateListFilesystemsError
+				p.listFilesystemsError = res.err
+				close(p.done)
+				continue
+			}
+
+			p.state = StateFanOutFilesystems
+
+			p.fsPruners = make([]*FSPruner, len(res.filesystems))
+			_, add, end := trace.WithTaskGroup(ctx, "pruner-fan-out-fs")
+			for i, fs := range res.filesystems {
+				p.fsPruners[i] = NewFSPruner(p.jid, p.side, p.keepRules, fs)
+				add(func(ctx context.Context) {
+					p.fsPruners[i].Run(ctx)
+				})
+			}
+			go func() {
+				end()
+				close(p.done)
+			}()
+
+		case req := <-p.stopReqs:
+			cancel()
+			go func() {
+				<-p.done
+				close(req.complete)
+			}()
+		case req := <-p.reportReqs:
+			req.reply <- p.report(req.ctx)
+		case <-p.done:
+			p.state = StateDone
+			return p.report(ctx)
+		}
+	}
+}
+
+func (p *Pruner) Report(ctx context.Context) *Report {
+	req := reportRequest{
+		ctx:   ctx,
+		reply: make(chan *Report, 1),
+	}
+	select {
+	case p.reportReqs <- req:
+		return <-req.reply
+	case <-ctx.Done():
+		return nil
+	case <-p.done:
+		return nil
+	}
+}
+
+func (p *Pruner) report(ctx context.Context) *Report {
+	fsreports := make([]*FSReport, len(p.fsPruners))
+	for i := range fsreports {
+		fsreports[i] = p.fsPruners[i].report()
+	}
+	return &Report{
+		State:                p.state,
+		ListFilesystemsError: p.listFilesystemsError,
+		Filesystems:          fsreports,
+	}
+}
+
+// implements pruning.Snapshot
+type snapshot struct {
+	replicated bool
+	stepHolds  []pruning.StepHold
+	zfs.FilesystemVersion
+
+	state     SnapState
+	destroyOp *zfs.DestroySnapOp
+}
+
+//go:generate enumer -type=SnapState -json
+type SnapState int
+
+const (
+	SnapStateInitialized SnapState = iota
+	SnapStateKeeping
+	SnapStateDeletePending
+	SnapStateDeleteAttempted
+)
+
+// implements pruning.StepHold
+type stepHold struct {
+	endpoint.Abstraction
+}
+
+func (s snapshot) Replicated() bool              { return s.replicated }
+func (s snapshot) StepHolds() []pruning.StepHold { return s.stepHolds }
+
+func (s stepHold) GetJobID() endpoint.JobID { return *s.Abstraction.GetJobID() }
+
+type FSPruner struct {
+	jid       endpoint.JobID
+	side      Side
+	keepRules []pruning.KeepRule
+	fsp       *zfs.DatasetPath
+
+	state FSState
+
+	// all channels consumed by the run loop
+	planned    chan fsPlanRes
+	executed   chan fsExecuteRes
+	done       chan struct{}
+	reportReqs chan fsReportReq
+
+	keepList    []*snapshot // valid in FSStateExecuting and forward
+	destroyList []*snapshot // valid in FSStateExecuting and forward, field .destroyOp is invalid until FSStateExecuting is left
+
+}
+
+type fsPlanRes struct {
+	keepList    []*snapshot
+	destroyList []*snapshot
+	err         error
+}
+
+type fsExecuteRes struct {
+	completedDestroyOps []*zfs.DestroySnapOp // same len() as FSPruner.destroyList
+}
+
+type fsReportReq struct {
+	res chan *FSReport
+}
+
+type FSReport struct {
+	State    FSState
+	KeepList []*SnapReport
+	Destroy  []*SnapReport
+}
+
+type SnapReport struct {
+	State         SnapState
+	Name          string
+	Replicated    bool
+	StepHoldCount int
+	DestroyError  error
+}
+
+//go:generate enumer -type=FSState -json
+type FSState int
+
+const (
+	FSStateInitialized FSState = iota
+	FSStatePlanning
+	FSStatePlanErr
+	FSStateExecuting
+	FSStateExecuteErr
+	FSStateExecuteSuccess
+)
+
+func (s FSState) IsTerminal() bool {
+	return s == FSStatePlanErr || s == FSStateExecuteErr || s == FSStateExecuteSuccess
+}
+
+func NewFSPruner(jid endpoint.JobID, side Side, keepRules []pruning.KeepRule, fsp *zfs.DatasetPath) *FSPruner {
+	return &FSPruner{
+		jid, side, keepRules, fsp,
+		FSStateInitialized,
+		make(chan fsPlanRes),
+		make(chan fsExecuteRes),
+		make(chan struct{}),
+		make(chan fsReportReq),
+		nil, nil,
+	}
+}
+
+func (p *FSPruner) Run(ctx context.Context) *FSReport {
+
+	defer func() {
+	}()
+
+	p.state = FSStatePlanning
+
+	go func() { p.planned <- p.plan(ctx) }()
+
+out:
+	for !p.state.IsTerminal() {
+		select {
+		case res := <-p.planned:
+
+			if res.err != nil {
+				p.state = FSStatePlanErr
+				continue
+			}
+			p.state = FSStateExecuting
+			p.keepList = res.keepList
+			p.destroyList = res.destroyList
+
+			go func() { p.executed <- p.execute(ctx, p.destroyList) }()
+
+		case res := <-p.executed:
+
+			if len(res.completedDestroyOps) != len(p.destroyList) {
+				panic("impl error: completedDestroyOps is a vector corresponding to entries in p.destroyList")
+			}
+
+			var erronous []*zfs.DestroySnapOp
+			for i, op := range res.completedDestroyOps {
+				if *op.ErrOut != nil {
+					erronous = append(erronous, op)
+				}
+				p.destroyList[i].destroyOp = op
+				p.destroyList[i].state = SnapStateDeleteAttempted
+			}
+			if len(erronous) > 0 {
+				p.state = FSStateExecuteErr
+			} else {
+				p.state = FSStateExecuteSuccess
+			}
+
+			close(p.done)
+
+		case <-p.reportReqs:
+			panic("unimp")
+		case <-p.done:
+			break out
+		}
+	}
+
+	// TODO render last FS report
+	return nil
+}
+
+func (p *FSPruner) plan(ctx context.Context) fsPlanRes {
+	fs := p.fsp.ToString()
+	vs, err := zfs.ZFSListFilesystemVersions(ctx, p.fsp, zfs.ListFilesystemVersionsOptions{})
+	if err != nil {
+		return fsPlanRes{err: errors.Wrap(err, "list filesystem versions")}
+	}
+
+	allJobsStepHolds, absErrs, err := endpoint.ListAbstractions(ctx, endpoint.ListZFSHoldsAndBookmarksQuery{
+		FS: endpoint.ListZFSHoldsAndBookmarksQueryFilesystemFilter{
+			FS: &fs,
+		},
+		What: endpoint.AbstractionTypeSet{
+			endpoint.AbstractionStepHold: true,
+		},
+		Concurrency: 1,
+	})
+	if err != nil {
+		return fsPlanRes{err: errors.Wrap(err, "list abstractions")}
+	}
+	if len(absErrs) > 0 {
+		logging.GetLogger(ctx, logging.SubsysPruning).WithError(endpoint.ListAbstractionsErrors(absErrs)).
+			Error("error listing some step holds, prune attempt might fail with 'dataset is busy' errors")
+	}
+
+	repPos, err := p.side.GetReplicationPosition(ctx, p.fsp.ToString())
+	if err != nil {
+		return fsPlanRes{err: errors.Wrap(err, "get replication position")}
+	}
+
+	vsAsSnaps := make([]pruning.Snapshot, len(vs))
+	for i := range vs {
+		var repPosCreateTxgOrZero uint64
+		if repPos != nil {
+			repPosCreateTxgOrZero = repPos.GetCreateTXG()
+		}
+		s := &snapshot{
+			state:             SnapStateInitialized,
+			FilesystemVersion: vs[i],
+			replicated:        vs[i].GetCreateTXG() <= repPosCreateTxgOrZero,
+		}
+		for _, h := range allJobsStepHolds {
+			if zfs.FilesystemVersionEqualIdentity(vs[i], h.GetFilesystemVersion()) {
+				s.stepHolds = append(s.stepHolds, stepHold{h})
+			}
+		}
+		vsAsSnaps[i] = s
+	}
+
+	downcastToSnapshots := func(l []pruning.Snapshot) (r []*snapshot) {
+		r = make([]*snapshot, len(l))
+		for i, e := range l {
+			r[i] = e.(*snapshot)
+		}
+		return r
+	}
+	pruningResult := pruning.PruneSnapshots(vsAsSnaps, p.keepRules)
+	remove, keep := downcastToSnapshots(pruningResult.Remove), downcastToSnapshots(pruningResult.Keep)
+	if len(remove)+len(keep) != len(vsAsSnaps) {
+		for _, s := range vsAsSnaps {
+			r, _ := json.MarshalIndent(s.(*snapshot).report(), "", "  ")
+			fmt.Fprintf(os.Stderr, "%s\n", string(r))
+		}
+		panic("indecisive")
+	}
+
+	for _, s := range remove {
+		s.state = SnapStateDeletePending
+	}
+	for _, s := range keep {
+		s.state = SnapStateKeeping
+	}
+
+	return fsPlanRes{keepList: keep, destroyList: remove, err: nil}
+}
+
+func (p *FSPruner) execute(ctx context.Context, destroyList []*snapshot) fsExecuteRes {
+	ops := make([]*zfs.DestroySnapOp, len(destroyList))
+	for i, fsv := range p.destroyList {
+		ops[i] = &zfs.DestroySnapOp{
+			Filesystem: p.fsp.ToString(),
+			Name:       fsv.GetName(),
+			ErrOut:     new(error),
+		}
+	}
+	zfs.ZFSDestroyFilesystemVersions(ctx, ops)
+
+	return fsExecuteRes{completedDestroyOps: ops}
+}
+
+func (p *FSPruner) report() *FSReport {
+	return &FSReport{
+		State:    p.state,
+		KeepList: p.reportRenderSnapReports(p.keepList),
+		Destroy:  p.reportRenderSnapReports(p.destroyList),
+	}
+}
+
+func (p *FSPruner) reportRenderSnapReports(l []*snapshot) (r []*SnapReport) {
+	r = make([]*SnapReport, len(l))
+	for i := range l {
+		r[i] = l[i].report()
+	}
+	return r
+}
+
+func (s *snapshot) report() *SnapReport {
+	var snapErr error
+	if s.state == SnapStateDeleteAttempted {
+		if *s.destroyOp.ErrOut != nil {
+			snapErr = (*s.destroyOp.ErrOut)
+		}
+	}
+	return &SnapReport{
+		State:         s.state,
+		Name:          s.Name,
+		Replicated:    s.Replicated(),
+		StepHoldCount: len(s.stepHolds),
+		DestroyError:  snapErr,
+	}
+}

daemon/pruner.v2/pruner_side.go

@@ -0,0 +1,27 @@
+package pruner
+
+import (
+	"context"
+
+	"github.com/zrepl/zrepl/endpoint"
+	"github.com/zrepl/zrepl/zfs"
+)
+
+type SideSender struct {
+	jobID endpoint.JobID
+}
+
+func NewSideSender(jid endpoint.JobID) *SideSender {
+	return &SideSender{jid}
+}
+
+func (s *SideSender) isSide() Side { return nil }
+
+var _ Side = (*SideSender)(nil)
+
+func (s *SideSender) GetReplicationPosition(ctx context.Context, fs string) (*zfs.FilesystemVersion, error) {
+	if fs == "" {
+		panic("must not pass zero value for fs")
+	}
+	return endpoint.GetMostRecentReplicationCursorOfJob(ctx, fs, s.jobID)
+}

daemon/pruner.v2/snapstate_enumer.go

@@ -0,0 +1,70 @@
+// Code generated by "enumer -type=SnapState -json"; DO NOT EDIT.
+
+//
+package pruner
+
+import (
+	"encoding/json"
+	"fmt"
+)
+
+const _SnapStateName = "SnapStateInitializedSnapStateKeepingSnapStateDeletePendingSnapStateDeleteAttempted"
+
+var _SnapStateIndex = [...]uint8{0, 20, 36, 58, 82}
+
+func (i SnapState) String() string {
+	if i < 0 || i >= SnapState(len(_SnapStateIndex)-1) {
+		return fmt.Sprintf("SnapState(%d)", i)
+	}
+	return _SnapStateName[_SnapStateIndex[i]:_SnapStateIndex[i+1]]
+}
+
+var _SnapStateValues = []SnapState{0, 1, 2, 3}
+
+var _SnapStateNameToValueMap = map[string]SnapState{
+	_SnapStateName[0:20]:  0,
+	_SnapStateName[20:36]: 1,
+	_SnapStateName[36:58]: 2,
+	_SnapStateName[58:82]: 3,
+}
+
+// SnapStateString retrieves an enum value from the enum constants string name.
+// Throws an error if the param is not part of the enum.
+func SnapStateString(s string) (SnapState, error) {
+	if val, ok := _SnapStateNameToValueMap[s]; ok {
+		return val, nil
+	}
+	return 0, fmt.Errorf("%s does not belong to SnapState values", s)
+}
+
+// SnapStateValues returns all values of the enum
+func SnapStateValues() []SnapState {
+	return _SnapStateValues
+}
+
+// IsASnapState returns "true" if the value is listed in the enum definition. "false" otherwise
+func (i SnapState) IsASnapState() bool {
+	for _, v := range _SnapStateValues {
+		if i == v {
+			return true
+		}
+	}
+	return false
+}
+
+// MarshalJSON implements the json.Marshaler interface for SnapState
+func (i SnapState) MarshalJSON() ([]byte, error) {
+	return json.Marshal(i.String())
+}
+
+// UnmarshalJSON implements the json.Unmarshaler interface for SnapState
+func (i *SnapState) UnmarshalJSON(data []byte) error {
+	var s string
+	if err := json.Unmarshal(data, &s); err != nil {
+		return fmt.Errorf("SnapState should be a string, got %s", data)
+	}
+
+	var err error
+	*i, err = SnapStateString(s)
+	return err
+}

daemon/pruner.v2/state_enumer.go

@@ -0,0 +1,71 @@
+// Code generated by "enumer -type=State -json"; DO NOT EDIT.
+
+//
+package pruner
+
+import (
+	"encoding/json"
+	"fmt"
+)
+
+const _StateName = "StateInitializedStateListFilesystemsStateListFilesystemsErrorStateFanOutFilesystemsStateDone"
+
+var _StateIndex = [...]uint8{0, 16, 36, 61, 83, 92}
+
+func (i State) String() string {
+	if i < 0 || i >= State(len(_StateIndex)-1) {
+		return fmt.Sprintf("State(%d)", i)
+	}
+	return _StateName[_StateIndex[i]:_StateIndex[i+1]]
+}
+
+var _StateValues = []State{0, 1, 2, 3, 4}
+
+var _StateNameToValueMap = map[string]State{
+	_StateName[0:16]:  0,
+	_StateName[16:36]: 1,
+	_StateName[36:61]: 2,
+	_StateName[61:83]: 3,
+	_StateName[83:92]: 4,
+}
+
+// StateString retrieves an enum value from the enum constants string name.
+// Throws an error if the param is not part of the enum.
+func StateString(s string) (State, error) {
+	if val, ok := _StateNameToValueMap[s]; ok {
+		return val, nil
+	}
+	return 0, fmt.Errorf("%s does not belong to State values", s)
+}
+
+// StateValues returns all values of the enum
+func StateValues() []State {
+	return _StateValues
+}
+
+// IsAState returns "true" if the value is listed in the enum definition. "false" otherwise
+func (i State) IsAState() bool {
+	for _, v := range _StateValues {
+		if i == v {
+			return true
+		}
+	}
+	return false
+}
+
+// MarshalJSON implements the json.Marshaler interface for State
+func (i State) MarshalJSON() ([]byte, error) {
+	return json.Marshal(i.String())
+}
+
+// UnmarshalJSON implements the json.Unmarshaler interface for State
+func (i *State) UnmarshalJSON(data []byte) error {
+	var s string
+	if err := json.Unmarshal(data, &s); err != nil {
+		return fmt.Errorf("State should be a string, got %s", data)
+	}
+
+	var err error
+	*i, err = StateString(s)
+	return err
+}

daemon/pruner/pruner.go

@@ -12,6 +12,7 @@ import (
 	"github.com/prometheus/client_golang/prometheus"
 
 	"github.com/zrepl/zrepl/config"
+	"github.com/zrepl/zrepl/daemon/logging"
 	"github.com/zrepl/zrepl/logger"
 	"github.com/zrepl/zrepl/pruning"
 	"github.com/zrepl/zrepl/replication/logic/pdu"
@@ -19,15 +20,14 @@ import (
 )
 
 // Try to keep it compatible with github.com/zrepl/zrepl/endpoint.Endpoint
-type History interface {
-	ReplicationCursor(ctx context.Context, req *pdu.ReplicationCursorReq) (*pdu.ReplicationCursorRes, error)
+type Endpoint interface {
 	ListFilesystems(ctx context.Context, req *pdu.ListFilesystemReq) (*pdu.ListFilesystemRes, error)
+	ListFilesystemVersions(ctx context.Context, req *pdu.ListFilesystemVersionsReq) (*pdu.ListFilesystemVersionsRes, error)
 }
 
 // Try to keep it compatible with github.com/zrepl/zrepl/endpoint.Endpoint
 type Target interface {
-	ListFilesystems(ctx context.Context, req *pdu.ListFilesystemReq) (*pdu.ListFilesystemRes, error)
-	ListFilesystemVersions(ctx context.Context, req *pdu.ListFilesystemVersionsReq) (*pdu.ListFilesystemVersionsRes, error)
+	Endpoint
 	DestroySnapshots(ctx context.Context, req *pdu.DestroySnapshotsReq) (*pdu.DestroySnapshotsRes, error)
 }
 
@@ -35,27 +35,24 @@ type Logger = logger.Logger
 
 type contextKey int
 
-const contextKeyLogger contextKey = 0
-
-func WithLogger(ctx context.Context, log Logger) context.Context {
-	return context.WithValue(ctx, contextKeyLogger, log)
-}
+const (
+	contextKeyPruneSide contextKey = 1 + iota
+)
 
 func GetLogger(ctx context.Context) Logger {
-	if l, ok := ctx.Value(contextKeyLogger).(Logger); ok {
-		return l
-	}
-	return logger.NewNullLogger()
+	pruneSide := ctx.Value(contextKeyPruneSide).(string)
+	return logging.GetLogger(ctx, logging.SubsysPruning).WithField("prune_side", pruneSide)
 }
 
 type args struct {
-	ctx                            context.Context
-	target                         Target
-	receiver                       History
-	rules                          []pruning.KeepRule
-	retryWait                      time.Duration
-	considerSnapAtCursorReplicated bool
-	promPruneSecs                  prometheus.Observer
+	ctx                              context.Context
+	target                           Target
+	sender, receiver                 Endpoint
+	rules                            []pruning.KeepRule
+	retryWait                        time.Duration
+	considerSnapAtCursorReplicated   bool
+	convertAnyStepHoldToStepBookmark bool
+	promPruneSecs                    prometheus.Observer
 }
 
 type Pruner struct {
@@ -73,11 +70,12 @@ type Pruner struct {
 }
 
 type PrunerFactory struct {
-	senderRules                    []pruning.KeepRule
-	receiverRules                  []pruning.KeepRule
-	retryWait                      time.Duration
-	considerSnapAtCursorReplicated bool
-	promPruneSecs                  *prometheus.HistogramVec
+	senderRules                      []pruning.KeepRule
+	receiverRules                    []pruning.KeepRule
+	retryWait                        time.Duration
+	considerSnapAtCursorReplicated   bool
+	convertAnyStepHoldToStepBookmark bool
+	promPruneSecs                    *prometheus.HistogramVec
 }
 
 type LocalPrunerFactory struct {
@@ -125,25 +123,35 @@ func NewPrunerFactory(in config.PruningSenderReceiver, promPruneSecs *prometheus
 		}
 		considerSnapAtCursorReplicated = considerSnapAtCursorReplicated || !knr.KeepSnapshotAtCursor
 	}
+
+	convertAnyStepHoldToStepBookmark := false
+	for _, r := range in.KeepSender {
+		_, ok := r.Ret.(*config.PruneKeepStepHolds)
+		convertAnyStepHoldToStepBookmark = convertAnyStepHoldToStepBookmark || ok
+	}
+
 	f := &PrunerFactory{
-		senderRules:                    keepRulesSender,
-		receiverRules:                  keepRulesReceiver,
-		retryWait:                      envconst.Duration("ZREPL_PRUNER_RETRY_INTERVAL", 10*time.Second),
-		considerSnapAtCursorReplicated: considerSnapAtCursorReplicated,
-		promPruneSecs:                  promPruneSecs,
+		senderRules:                      keepRulesSender,
+		receiverRules:                    keepRulesReceiver,
+		retryWait:                        envconst.Duration("ZREPL_PRUNER_RETRY_INTERVAL", 10*time.Second),
+		considerSnapAtCursorReplicated:   considerSnapAtCursorReplicated,
+		convertAnyStepHoldToStepBookmark: convertAnyStepHoldToStepBookmark,
+		promPruneSecs:                    promPruneSecs,
 	}
 	return f, nil
 }
 
-func (f *PrunerFactory) BuildSenderPruner(ctx context.Context, target Target, receiver History) *Pruner {
+func (f *PrunerFactory) BuildSenderPruner(ctx context.Context, sender Target, receiver Endpoint) *Pruner {
 	p := &Pruner{
 		args: args{
-			WithLogger(ctx, GetLogger(ctx).WithField("prune_side", "sender")),
-			target,
+			context.WithValue(ctx, contextKeyPruneSide, "sender"),
+			sender,
+			sender,
 			receiver,
 			f.senderRules,
 			f.retryWait,
 			f.considerSnapAtCursorReplicated,
+			f.convertAnyStepHoldToStepBookmark,
 			f.promPruneSecs.WithLabelValues("sender"),
 		},
 		state: Plan,
@@ -151,15 +159,17 @@ func (f *PrunerFactory) BuildSenderPruner(ctx context.Context, target Target, re
 	return p
 }
 
-func (f *PrunerFactory) BuildReceiverPruner(ctx context.Context, target Target, receiver History) *Pruner {
+func (f *PrunerFactory) BuildReceiverPruner(ctx context.Context, receiver Target, sender Endpoint) *Pruner {
 	p := &Pruner{
 		args: args{
-			WithLogger(ctx, GetLogger(ctx).WithField("prune_side", "receiver")),
-			target,
+			context.WithValue(ctx, contextKeyPruneSide, "receiver"),
+			receiver,
+			sender,
 			receiver,
 			f.receiverRules,
 			f.retryWait,
 			false, // senseless here anyways
+			false, // senseless here anyways
 			f.promPruneSecs.WithLabelValues("receiver"),
 		},
 		state: Plan,
@@ -167,15 +177,17 @@ func (f *PrunerFactory) BuildReceiverPruner(ctx context.Context, target Target,
 	return p
 }
 
-func (f *LocalPrunerFactory) BuildLocalPruner(ctx context.Context, target Target, receiver History) *Pruner {
+func (f *LocalPrunerFactory) BuildLocalPruner(ctx context.Context, target Target) *Pruner {
 	p := &Pruner{
 		args: args{
-			ctx,
+			context.WithValue(ctx, contextKeyPruneSide, "local"),
+			target,
+			target,
 			target,
-			receiver,
 			f.keepRules,
 			f.retryWait,
 			false, // considerSnapAtCursorReplicated is not relevant for local pruning
+			false, // convertAnyStepHoldToStepBookmark is not relevant for local pruning
 			f.promPruneSecs.WithLabelValues("local"),
 		},
 		state: Plan,
@@ -344,11 +356,13 @@ func (s snapshot) Replicated() bool { return s.replicated }
 
 func (s snapshot) Date() time.Time { return s.date }
 
+func (s snapshot) CreateTXG() uint64 { return s.fsv.GetCreateTXG() }
+
 func doOneAttempt(a *args, u updater) {
 
-	ctx, target, receiver := a.ctx, a.target, a.receiver
+	ctx, sender, receiver, target := a.ctx, a.sender, a.receiver, a.target
 
-	sfssres, err := receiver.ListFilesystems(ctx, &pdu.ListFilesystemReq{})
+	sfssres, err := sender.ListFilesystems(ctx, &pdu.ListFilesystemReq{})
 	if err != nil {
 		u(func(p *Pruner) {
 			p.state = PlanErr
@@ -410,6 +424,10 @@ tfss_loop:
 
 		pfs.snaps = make([]pruning.Snapshot, 0, len(tfsvs))
 
+		receiver.ListFilesystemVersions(ctx, &pdu.ListFilesystemVersionsReq{
+			Filesystem: tfs.Path,
+		})
+
 		rcReq := &pdu.ReplicationCursorReq{
 			Filesystem: tfs.Path,
 		}
@@ -418,6 +436,7 @@ tfss_loop:
 			pfsPlanErrAndLog(err, "cannot get replication cursor bookmark")
 			continue tfss_loop
 		}
+
 		if rc.GetNotexist() {
 			err := errors.New("replication cursor bookmark does not exist (one successful replication is required before pruning works)")
 			pfsPlanErrAndLog(err, "")

daemon/snapper/snapper.go

@@ -8,10 +8,12 @@ import (
 	"time"
 
 	"github.com/pkg/errors"
+	"github.com/zrepl/zrepl/daemon/logging/trace"
 
 	"github.com/zrepl/zrepl/config"
 	"github.com/zrepl/zrepl/daemon/filters"
 	"github.com/zrepl/zrepl/daemon/hooks"
+	"github.com/zrepl/zrepl/daemon/logging"
 	"github.com/zrepl/zrepl/logger"
 	"github.com/zrepl/zrepl/util/envconst"
 	"github.com/zrepl/zrepl/zfs"
@@ -45,7 +47,6 @@ type snapProgress struct {
 
 type args struct {
 	ctx            context.Context
-	log            Logger
 	prefix         string
 	interval       time.Duration
 	fsf            *filters.DatasetMapFilter
@@ -102,23 +103,10 @@ func (s State) sf() state {
 type updater func(u func(*Snapper)) State
 type state func(a args, u updater) state
 
-type contextKey int
-
-const (
-	contextKeyLog contextKey = 0
-)
-
 type Logger = logger.Logger
 
-func WithLogger(ctx context.Context, log Logger) context.Context {
-	return context.WithValue(ctx, contextKeyLog, log)
-}
-
 func getLogger(ctx context.Context) Logger {
-	if log, ok := ctx.Value(contextKeyLog).(Logger); ok {
-		return log
-	}
-	return logger.NewNullLogger()
+	return logging.GetLogger(ctx, logging.SubsysSnapshot)
 }
 
 func PeriodicFromConfig(g *config.Global, fsf *filters.DatasetMapFilter, in *config.SnapshottingPeriodic) (*Snapper, error) {
@@ -146,13 +134,12 @@ func PeriodicFromConfig(g *config.Global, fsf *filters.DatasetMapFilter, in *con
 }
 
 func (s *Snapper) Run(ctx context.Context, snapshotsTaken chan<- struct{}) {
-
+	defer trace.WithSpanFromStackUpdateCtx(&ctx)()
 	getLogger(ctx).Debug("start")
 	defer getLogger(ctx).Debug("stop")
 
 	s.args.snapshotsTaken = snapshotsTaken
 	s.args.ctx = ctx
-	s.args.log = getLogger(ctx)
 	s.args.dryRun = false // for future expansion
 
 	u := func(u func(*Snapper)) State {
@@ -190,7 +177,7 @@ func onErr(err error, u updater) state {
 		case Snapshotting:
 			s.state = ErrorWait
 		}
-		s.args.log.WithError(err).WithField("pre_state", preState).WithField("post_state", s.state).Error("snapshotting error")
+		getLogger(s.args.ctx).WithError(err).WithField("pre_state", preState).WithField("post_state", s.state).Error("snapshotting error")
 	}).sf()
 }
 
@@ -209,7 +196,7 @@ func syncUp(a args, u updater) state {
 	if err != nil {
 		return onErr(err, u)
 	}
-	syncPoint, err := findSyncPoint(a.log, fss, a.prefix, a.interval)
+	syncPoint, err := findSyncPoint(a.ctx, fss, a.prefix, a.interval)
 	if err != nil {
 		return onErr(err, u)
 	}
@@ -266,18 +253,18 @@ func snapshot(a args, u updater) state {
 		suffix := time.Now().In(time.UTC).Format("20060102_150405_000")
 		snapname := fmt.Sprintf("%s%s", a.prefix, suffix)
 
-		l := a.log.
-			WithField("fs", fs.ToString()).
-			WithField("snap", snapname)
+		ctx := logging.WithInjectedField(a.ctx, "fs", fs.ToString())
+		ctx = logging.WithInjectedField(ctx, "snap", snapname)
 
 		hookEnvExtra := hooks.Env{
 			hooks.EnvFS:       fs.ToString(),
 			hooks.EnvSnapshot: snapname,
 		}
 
-		jobCallback := hooks.NewCallbackHookForFilesystem("snapshot", fs, func(_ context.Context) (err error) {
+		jobCallback := hooks.NewCallbackHookForFilesystem("snapshot", fs, func(ctx context.Context) (err error) {
+			l := getLogger(ctx)
 			l.Debug("create snapshot")
-			err = zfs.ZFSSnapshot(a.ctx, fs, snapname, false) // TODO propagate context to ZFSSnapshot
+			err = zfs.ZFSSnapshot(ctx, fs, snapname, false) // TODO propagate context to ZFSSnapshot
 			if err != nil {
 				l.WithError(err).Error("cannot create snapshot")
 			}
@@ -290,7 +277,7 @@ func snapshot(a args, u updater) state {
 		{
 			filteredHooks, err := a.hooks.CopyFilteredForFilesystem(fs)
 			if err != nil {
-				l.WithError(err).Error("unexpected filter error")
+				getLogger(ctx).WithError(err).Error("unexpected filter error")
 				fsHadErr = true
 				goto updateFSState
 			}
@@ -303,7 +290,7 @@ func snapshot(a args, u updater) state {
 			plan, planErr = hooks.NewPlan(&filteredHooks, hooks.PhaseSnapshot, jobCallback, hookEnvExtra)
 			if planErr != nil {
 				fsHadErr = true
-				l.WithError(planErr).Error("cannot create job hook plan")
+				getLogger(ctx).WithError(planErr).Error("cannot create job hook plan")
 				goto updateFSState
 			}
 		}
@@ -314,15 +301,14 @@ func snapshot(a args, u updater) state {
 			progress.state = SnapStarted
 		})
 		{
-			l := hooks.GetLogger(a.ctx).WithField("fs", fs.ToString()).WithField("snap", snapname)
-			l.WithField("report", plan.Report().String()).Debug("begin run job plan")
-			plan.Run(hooks.WithLogger(a.ctx, l), a.dryRun)
+			getLogger(ctx).WithField("report", plan.Report().String()).Debug("begin run job plan")
+			plan.Run(ctx, a.dryRun)
 			planReport = plan.Report()
 			fsHadErr = planReport.HadError() // not just fatal errors
 			if fsHadErr {
-				l.WithField("report", planReport.String()).Error("end run job plan with error")
+				getLogger(ctx).WithField("report", planReport.String()).Error("end run job plan with error")
 			} else {
-				l.WithField("report", planReport.String()).Info("end run job plan successful")
+				getLogger(ctx).WithField("report", planReport.String()).Info("end run job plan successful")
 			}
 		}
 
@@ -342,7 +328,7 @@ func snapshot(a args, u updater) state {
 	case a.snapshotsTaken <- struct{}{}:
 	default:
 		if a.snapshotsTaken != nil {
-			a.log.Warn("callback channel is full, discarding snapshot update event")
+			getLogger(a.ctx).Warn("callback channel is full, discarding snapshot update event")
 		}
 	}
 
@@ -355,7 +341,7 @@ func snapshot(a args, u updater) state {
 					break
 				}
 			}
-			a.log.WithField("hook", h.String()).WithField("hook_number", hookIdx+1).Warn("hook did not match any snapshotted filesystems")
+			getLogger(a.ctx).WithField("hook", h.String()).WithField("hook_number", hookIdx+1).Warn("hook did not match any snapshotted filesystems")
 		}
 	}
 
@@ -376,7 +362,7 @@ func wait(a args, u updater) state {
 		lastTick := snapper.lastInvocation
 		snapper.sleepUntil = lastTick.Add(a.interval)
 		sleepUntil = snapper.sleepUntil
-		log := a.log.WithField("sleep_until", sleepUntil).WithField("duration", a.interval)
+		log := getLogger(a.ctx).WithField("sleep_until", sleepUntil).WithField("duration", a.interval)
 		logFunc := log.Debug
 		if snapper.state == ErrorWait || snapper.state == SyncUpErrWait {
 			logFunc = log.Error
@@ -404,7 +390,7 @@ func listFSes(ctx context.Context, mf *filters.DatasetMapFilter) (fss []*zfs.Dat
 var syncUpWarnNoSnapshotUntilSyncupMinDuration = envconst.Duration("ZREPL_SNAPPER_SYNCUP_WARN_MIN_DURATION", 1*time.Second)
 
 // see docs/snapshotting.rst
-func findSyncPoint(log Logger, fss []*zfs.DatasetPath, prefix string, interval time.Duration) (syncPoint time.Time, err error) {
+func findSyncPoint(ctx context.Context, fss []*zfs.DatasetPath, prefix string, interval time.Duration) (syncPoint time.Time, err error) {
 
 	const (
 		prioHasVersions int = iota
@@ -426,10 +412,10 @@ func findSyncPoint(log Logger, fss []*zfs.DatasetPath, prefix string, interval t
 
 	now := time.Now()
 
-	log.Debug("examine filesystem state to find sync point")
+	getLogger(ctx).Debug("examine filesystem state to find sync point")
 	for _, d := range fss {
-		l := log.WithField("fs", d.ToString())
-		syncPoint, err := findSyncPointFSNextOptimalSnapshotTime(l, now, interval, prefix, d)
+		ctx := logging.WithInjectedField(ctx, "fs", d.ToString())
+		syncPoint, err := findSyncPointFSNextOptimalSnapshotTime(ctx, now, interval, prefix, d)
 		if err == findSyncPointFSNoFilesystemVersionsErr {
 			snaptimes = append(snaptimes, snapTime{
 				ds:   d,
@@ -438,9 +424,9 @@ func findSyncPoint(log Logger, fss []*zfs.DatasetPath, prefix string, interval t
 			})
 		} else if err != nil {
 			hardErrs++
-			l.WithError(err).Error("cannot determine optimal sync point for this filesystem")
+			getLogger(ctx).WithError(err).Error("cannot determine optimal sync point for this filesystem")
 		} else {
-			l.WithField("syncPoint", syncPoint).Debug("found optimal sync point for this filesystem")
+			getLogger(ctx).WithField("syncPoint", syncPoint).Debug("found optimal sync point for this filesystem")
 			snaptimes = append(snaptimes, snapTime{
 				ds:   d,
 				prio: prioHasVersions,
@@ -467,7 +453,7 @@ func findSyncPoint(log Logger, fss []*zfs.DatasetPath, prefix string, interval t
 	})
 
 	winnerSyncPoint := snaptimes[0].time
-	l := log.WithField("syncPoint", winnerSyncPoint.String())
+	l := getLogger(ctx).WithField("syncPoint", winnerSyncPoint.String())
 	l.Info("determined sync point")
 	if winnerSyncPoint.Sub(now) > syncUpWarnNoSnapshotUntilSyncupMinDuration {
 		for _, st := range snaptimes {
@@ -483,9 +469,9 @@ func findSyncPoint(log Logger, fss []*zfs.DatasetPath, prefix string, interval t
 
 var findSyncPointFSNoFilesystemVersionsErr = fmt.Errorf("no filesystem versions")
 
-func findSyncPointFSNextOptimalSnapshotTime(l Logger, now time.Time, interval time.Duration, prefix string, d *zfs.DatasetPath) (time.Time, error) {
+func findSyncPointFSNextOptimalSnapshotTime(ctx context.Context, now time.Time, interval time.Duration, prefix string, d *zfs.DatasetPath) (time.Time, error) {
 
-	fsvs, err := zfs.ZFSListFilesystemVersions(d, zfs.ListFilesystemVersionsOptions{
+	fsvs, err := zfs.ZFSListFilesystemVersions(ctx, d, zfs.ListFilesystemVersionsOptions{
 		Types:           zfs.Snapshots,
 		ShortnamePrefix: prefix,
 	})
@@ -502,7 +488,7 @@ func findSyncPointFSNextOptimalSnapshotTime(l Logger, now time.Time, interval ti
 	})
 
 	latest := fsvs[len(fsvs)-1]
-	l.WithField("creation", latest.Creation).Debug("found latest snapshot")
+	getLogger(ctx).WithField("creation", latest.Creation).Debug("found latest snapshot")
 
 	since := now.Sub(latest.Creation)
 	if since < 0 {

docs/changelog.rst

@@ -62,7 +62,9 @@ Actual changelog:
 * |bugfix| |docs| snapshotting: clarify sync-up behavior and warn about filesystems
   that will not be snapshotted until the sync-up phase is over
 * |docs| Document new replication features in the :ref:`config overview <overview-how-replication-works>` and :repomasterlink:`replication/design.md`.
+* |feature| documented subcommand to generate ``bash`` and ``zsh`` completions
 * **[MAINTAINER NOTICE]** New platform tests in this version, please make sure you run them for your distro!
+* **[MAINTAINER NOTICE]** Please add the shell completions to the zrepl packages.
 
 0.2.1
 -----

docs/configuration/transports.rst

@@ -52,8 +52,15 @@ Serve
         listen: ":8888"
         listen_freebind: true # optional, default false
         clients: {
-          "192.168.122.123" : "mysql01"
-          "192.168.122.123" : "mx01"
+          "192.168.122.123" :               "mysql01",
+          "192.168.122.42" :                "mx01",
+          "2001:0db8:85a3::8a2e:0370:7334": "gateway",
+
+          # CIDR masks require a '*' in the client identity string
+          # that is expanded to the client's IP address
+
+          "10.23.42.0/24":       "cluster-*"
+          "fde4:8dba:82e1::/64": "san-*"
         }
       ...
 

docs/installation.rst

@@ -1,138 +1,19 @@
-.. _binary releases: https://github.com/zrepl/zrepl/releases
 
-.. _installation:
+.. _installation_toc:
 
+************
 Installation
-============
+************
 
 .. TIP::
 
     Note: check out the :ref:`tutorial` if you want a first impression of zrepl.
 
-User Privileges
----------------
+.. toctree::
 
-It is possible to run zrepl as an unprivileged user in combination with
-`ZFS delegation <https://www.freebsd.org/doc/handbook/zfs-zfs-allow.html>`_.
-Also, there is the possibility to run it in a jail on FreeBSD by delegating a dataset to the jail.
-However, until we get around documenting those setups, you will have to run zrepl as root or experiment yourself :)
-
-Packages
---------
-
-zrepl source releases are signed & tagged by the author in the git repository.
-Your OS vendor may provide binary packages of zrepl through the package manager.
-Additionally, `binary releases`_ are provided on GitHub.
-The following list may be incomplete, feel free to submit a PR with an update:
-
-.. list-table::
-    :header-rows: 1
-
-    * - OS / Distro
-      - Install Command
-      - Link
-    * - FreeBSD
-      - ``pkg install zrepl``
-      - `<https://www.freshports.org/sysutils/zrepl/>`_
-    * - MacOS
-      - ``brew install zrepl``
-      - Available on `homebrew <https://brew.sh>`_
-    * - Arch Linux
-      - ``yay install zrepl``
-      - Available on `AUR <https://aur.archlinux.org/packages/zrepl>`_
-    * - Fedora
-      - ``dnf install zrepl``
-      - Available on `COPR <https://copr.fedorainfracloud.org/coprs/poettlerric/zrepl/>`_
-    * - CentOS/RHEL
-      - ``yum install zrepl``
-      - Available on `COPR <https://copr.fedorainfracloud.org/coprs/poettlerric/zrepl/>`_
-    * - Debian + Ubuntu
-      - ``apt install zrepl``
-      - APT repository config :ref:`see below <installation-apt-repos>`
-    * - OmniOS
-      - ``pkg install zrepl``
-      - Available since `r151030 <https://pkg.omniosce.org/r151030/extra/en/search.shtml?token=zrepl&action=Search>`_
-    * - Void Linux
-      - ``xbps-install zrepl``
-      - Available since `a88a2a4 <https://github.com/void-linux/void-packages/commit/a88a2a4d7bf56072dadf61ab56b8424e39155890>`_
-    * - Others
-      -
-      - Use `binary releases`_ or build from source.
-
-.. _installation-apt-repos:
-
-Debian / Ubuntu APT repositories
-~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
-
-We maintain APT repositories for Debian, Ubuntu and derivatives.
-The fingerprint of the signing key is ``E101 418F D3D6 FBCB 9D65  A62D 7086 99FC 5F2E BF16``.
-It is available at `<https://zrepl.cschwarz.com/apt/apt-key.asc>`_ .
-Please open an issue `in the packaging repository <https://github.com/zrepl/debian-binary-packaging>`_ if you encounter any issues with the repository.
-
-The following snippet configure the repository for your Debian or Ubuntu release:
-
-::
-
-    apt update && apt install curl gnupg lsb-release; \
-    ARCH="$(dpkg --print-architecture)"; \
-    CODENAME="$(lsb_release -i -s | tr '[:upper:]' '[:lower:]') $(lsb_release -c -s | tr '[:upper:]' '[:lower:]')"; \
-    echo "Using Distro and Codename: $CODENAME"; \
-    (curl https://zrepl.cschwarz.com/apt/apt-key.asc | apt-key add -) && \
-    (echo "deb [arch=$ARCH] https://zrepl.cschwarz.com/apt/$CODENAME main" > /etc/apt/sources.list.d/zrepl.list) && \
-    apt update
-
-
-.. NOTE::
-
-   Until zrepl reaches 1.0, all APT repositories will be updated to the latest zrepl release immediately.
-   This includes breaking changes between zrepl versions.
-   Use ``apt-mark hold zrepl`` to prevent upgrades of zrepl.
-
-Compile From Source
-~~~~~~~~~~~~~~~~~~~
-
-Producing a release requires **Go 1.11** or newer and **Python 3** + **pip3** + ``docs/requirements.txt`` for the Sphinx documentation.
-A tutorial to install Go is available over at `golang.org <https://golang.org/doc/install>`_.
-Python and pip3 should probably be installed via your distro's package manager.
-
-Alternatively, you can use the Docker build process:
-it is used to produce the official zrepl `binary releases`_
-and serves as a reference for build dependencies and procedure:
-
-::
-
-    git clone https://github.com/zrepl/zrepl.git && \
-    cd zrepl && \
-    sudo docker build -t zrepl_build -f build.Dockerfile . && \
-    sudo docker run -it --rm \
-        -v "${PWD}:/src" \
-        --user "$(id -u):$(id -g)" \
-        zrepl_build make release
-
-Alternatively, you can install build dependencies on your local system and then build in your ``$GOPATH``:
-
-::
-
-    mkdir -p "${GOPATH}/src/github.com/zrepl/zrepl"
-    git clone https://github.com/zrepl/zrepl.git "${GOPATH}/src/github.com/zrepl/zrepl"
-    cd "${GOPATH}/src/github.com/zrepl/zrepl"
-    python3 -m venv3
-    source venv3/bin/activate
-    ./lazy.sh devsetup
-    make release
-
-The Python venv is used for the documentation build dependencies.
-If you just want to build the zrepl binary, leave it out and use `./lazy.sh godep` instead.
-Either way, all build results are located in the ``artifacts/`` directory.
-
-.. NOTE::
-
-    It is your job to install the appropriate binary in the zrepl users's ``$PATH``, e.g. ``/usr/local/bin/zrepl``.
-    Otherwise, the examples in the :ref:`tutorial` may need to be adjusted.
-
-What next?
-----------
-
-Read the :ref:`configuration chapter<configuration_toc>` and then continue with the :ref:`usage chapter<usage>`.
-
-**Reminder**: If you want a quick introduction, please read the :ref:`tutorial`.
+    installation/user-privileges
+    installation/packages
+    installation/apt-repos
+    installation/compile-from-source
+    installation/freebsd-jail-with-iocage
+    installation/what-next

docs/installation/apt-repos.rst

@@ -0,0 +1,29 @@
+
+.. _installation-apt-repos:
+
+Debian / Ubuntu APT repositories
+~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
+
+We maintain APT repositories for Debian, Ubuntu and derivatives.
+The fingerprint of the signing key is ``E101 418F D3D6 FBCB 9D65  A62D 7086 99FC 5F2E BF16``.
+It is available at `<https://zrepl.cschwarz.com/apt/apt-key.asc>`_ .
+Please open an issue `in the packaging repository <https://github.com/zrepl/debian-binary-packaging>`_ if you encounter any issues with the repository.
+
+The following snippet configure the repository for your Debian or Ubuntu release:
+
+::
+
+    apt update && apt install curl gnupg lsb-release; \
+    ARCH="$(dpkg --print-architecture)"; \
+    CODENAME="$(lsb_release -i -s | tr '[:upper:]' '[:lower:]') $(lsb_release -c -s | tr '[:upper:]' '[:lower:]')"; \
+    echo "Using Distro and Codename: $CODENAME"; \
+    (curl https://zrepl.cschwarz.com/apt/apt-key.asc | apt-key add -) && \
+    (echo "deb [arch=$ARCH] https://zrepl.cschwarz.com/apt/$CODENAME main" > /etc/apt/sources.list.d/zrepl.list) && \
+    apt update
+
+
+.. NOTE::
+
+   Until zrepl reaches 1.0, all APT repositories will be updated to the latest zrepl release immediately.
+   This includes breaking changes between zrepl versions.
+   Use ``apt-mark hold zrepl`` to prevent upgrades of zrepl.

docs/installation/compile-from-source.rst

@@ -0,0 +1,45 @@
+.. _binary releases: https://github.com/zrepl/zrepl/releases
+
+.. _installation-compile-from-source:
+
+Compile From Source
+~~~~~~~~~~~~~~~~~~~
+
+Producing a release requires **Go 1.11** or newer and **Python 3** + **pip3** + ``docs/requirements.txt`` for the Sphinx documentation.
+A tutorial to install Go is available over at `golang.org <https://golang.org/doc/install>`_.
+Python and pip3 should probably be installed via your distro's package manager.
+
+Alternatively, you can use the Docker build process:
+it is used to produce the official zrepl `binary releases`_
+and serves as a reference for build dependencies and procedure:
+
+::
+
+    git clone https://github.com/zrepl/zrepl.git && \
+    cd zrepl && \
+    sudo docker build -t zrepl_build -f build.Dockerfile . && \
+    sudo docker run -it --rm \
+        -v "${PWD}:/src" \
+        --user "$(id -u):$(id -g)" \
+        zrepl_build make release
+
+Alternatively, you can install build dependencies on your local system and then build in your ``$GOPATH``:
+
+::
+
+    mkdir -p "${GOPATH}/src/github.com/zrepl/zrepl"
+    git clone https://github.com/zrepl/zrepl.git "${GOPATH}/src/github.com/zrepl/zrepl"
+    cd "${GOPATH}/src/github.com/zrepl/zrepl"
+    python3 -m venv3
+    source venv3/bin/activate
+    ./lazy.sh devsetup
+    make release
+
+The Python venv is used for the documentation build dependencies.
+If you just want to build the zrepl binary, leave it out and use `./lazy.sh godep` instead.
+Either way, all build results are located in the ``artifacts/`` directory.
+
+.. NOTE::
+
+    It is your job to install the appropriate binary in the zrepl users's ``$PATH``, e.g. ``/usr/local/bin/zrepl``.
+    Otherwise, the examples in the :ref:`tutorial` may need to be adjusted.

docs/installation/freebsd-jail-with-iocage.rst

@@ -0,0 +1,140 @@
+.. include:: ../global.rst.inc
+
+.. _installation-freebsd-jail-with-iocage:
+
+FreeBSD Jail With iocage
+========================
+
+
+This tutorial shows how zrepl can be installed on FreeBSD, or FreeNAS in a jail using iocage.
+While this tutorial focuses on using iocage, much of the setup would be similar
+using a different jail manager.
+
+.. NOTE::
+
+   From a security perspective, just keep in mind that ``zfs send``/``recv`` was never designed with
+   jails in mind, an attacker could probably crash the receive-side kernel or worse induce stateful
+   damage to the receive-side pool if they were able to get access to the jail.
+
+   The jail doesn't provide security benefits, but only management ones.
+
+Requirements
+------------
+
+A dataset that will be delegated to the jail needs to be created if one does not already exist.
+For the tutorial ``tank/zrepl`` will be used.
+
+.. code-block:: bash
+
+   zfs create -o mountpoint=none tank/zrepl
+
+The only software requirements on the host system are ``iocage``, which can be installed
+from ports or packages.
+
+.. code-block:: bash
+
+   pkg install py37-iocage
+
+.. NOTE::
+
+   By default ``iocage`` will "activate" on first use which will set up some defaults such as
+   which pool will be used. To activate ``iocage`` manually the ``iocage activate`` command can be used.
+
+Jail Creation
+-------------
+
+There are two options for jail creation using FreeBSD.
+
+1. Manually set up the jail from scratch
+2. Create the jail using the ``zrepl`` plugin. On FreeNAS this is possible from the user interface using the community index.
+
+Manual Jail
+###########
+
+Create a jail, using the same release as the host, called ``zrepl`` that will be automatically started at boot.
+The jail will have ``tank/zrepl`` delegated into it.
+
+.. code-block:: bash
+
+   iocage create --release "$(freebsd-version -k | cut -d '-' -f '1,2')" --name zrepl \
+          boot=on nat=1 \
+          jail_zfs=on \
+          jail_zfs_dataset=zrepl \
+          jail_zfs_mountpoint='none'
+
+Enter the jail:
+
+.. code-block:: bash
+
+   iocage console zrepl
+
+Install ``zrepl``
+
+.. code-block:: bash
+
+   pkg update && pkg upgrade
+   pkg install zrepl
+
+Create the log file ``/var/log/zrepl.log``
+
+.. code-block:: bash
+
+   touch /var/log/zrepl.log && service newsyslog restart
+
+Tell syslogd to redirect facility local0 to the ``zrepl.log`` file:
+
+.. code-block:: bash
+
+   service syslogd reload
+
+Enable the zrepl daemon to start automatically at boot:
+
+.. code-block:: bash
+
+   sysrc zrepl_enable="YES"
+
+
+Plugin
+######
+
+When using the plugin, ``zrepl`` will be installed for you in a jail using the following ``iocage`` properties.
+
+* ``nat=1``
+* ``jail_zfs=on``
+* ``jail_zfs_mountpoint=none``
+
+Additionally the delegated dataset should be specified upon creation, and optionally start on boot can be set.
+This can also be done from the FreeNAS webui.
+
+.. code-block:: bash
+
+   fetch https://raw.githubusercontent.com/ix-plugin-hub/iocage-plugin-index/master/zrepl.json -o /tmp/zrepl.json
+   iocage fetch -P /tmp/zrepl.json --name zrepl jail_zfs_dataset=zrepl boot=on
+
+Configuration
+-------------
+
+Now ``zrepl`` can be configured.
+
+Enter the jail.
+
+.. code-block:: bash
+
+   iocage console zrepl
+
+Modify the ``/usr/local/etc/zrepl/zrepl.yml`` configuration file.
+
+.. TIP::
+
+    Note: check out the :ref:`tutorial` for examples of a ``sink`` job.
+
+Now ``zrepl`` can be started.
+
+.. code-block:: bash
+
+   service zrepl start
+
+Summary
+-------
+
+Congratulations, you have a working jail!

docs/installation/packages.rst

@@ -0,0 +1,50 @@
+.. _installation-packages:
+
+.. _binary releases: https://github.com/zrepl/zrepl/releases
+
+Packages
+--------
+
+zrepl source releases are signed & tagged by the author in the git repository.
+Your OS vendor may provide binary packages of zrepl through the package manager.
+Additionally, `binary releases`_ are provided on GitHub.
+The following list may be incomplete, feel free to submit a PR with an update:
+
+.. list-table::
+    :header-rows: 1
+
+    * - OS / Distro
+      - Install Command
+      - Link
+    * - FreeBSD
+      - ``pkg install zrepl``
+      - `<https://www.freshports.org/sysutils/zrepl/>`_
+      
+        :ref:`installation-freebsd-jail-with-iocage`
+    * - FreeNAS
+      -
+      - :ref:`installation-freebsd-jail-with-iocage`
+    * - MacOS
+      - ``brew install zrepl``
+      - Available on `homebrew <https://brew.sh>`_
+    * - Arch Linux
+      - ``yay install zrepl``
+      - Available on `AUR <https://aur.archlinux.org/packages/zrepl>`_
+    * - Fedora
+      - ``dnf install zrepl``
+      - Available on `COPR <https://copr.fedorainfracloud.org/coprs/poettlerric/zrepl/>`_
+    * - CentOS/RHEL
+      - ``yum install zrepl``
+      - Available on `COPR <https://copr.fedorainfracloud.org/coprs/poettlerric/zrepl/>`_
+    * - Debian + Ubuntu
+      - ``apt install zrepl``
+      - APT repository config :ref:`see below <installation-apt-repos>`
+    * - OmniOS
+      - ``pkg install zrepl``
+      - Available since `r151030 <https://pkg.omniosce.org/r151030/extra/en/search.shtml?token=zrepl&action=Search>`_
+    * - Void Linux
+      - ``xbps-install zrepl``
+      - Available since `a88a2a4 <https://github.com/void-linux/void-packages/commit/a88a2a4d7bf56072dadf61ab56b8424e39155890>`_
+    * - Others
+      -
+      - Use `binary releases`_ or build from source.

docs/installation/user-privileges.rst

@@ -0,0 +1,12 @@
+.. _installation-user-privileges:
+
+User Privileges
+---------------
+
+It is possible to run zrepl as an unprivileged user in combination with
+`ZFS delegation <https://www.freebsd.org/doc/handbook/zfs-zfs-allow.html>`_.
+Also, there is the possibility to run it in a jail on FreeBSD by delegating a dataset to the jail.
+
+.. TIP::
+
+    Note: check out the :ref:`installation-freebsd-jail-with-iocage` for FreeBSD jail setup instructions.

docs/installation/what-next.rst

@@ -0,0 +1,8 @@
+.. _installation-what-next:
+
+What next?
+----------
+
+Read the :ref:`configuration chapter<configuration_toc>` and then continue with the :ref:`usage chapter<usage>`.
+
+**Reminder**: If you want a quick introduction, please read the :ref:`tutorial`.

docs/supporters.rst

@@ -32,6 +32,9 @@ We would like to thank the following people and organizations for supporting zre
 
    <div class="fa fa-code" style="width: 1em;"></div>
 
+* |supporter-std| Henning Kessler
+* |supporter-std| `John Ramsden <https://github.com/johnramsden>`_
+* |supporter-std| `DrLuke <https://github.com/drluke>`_
 * |supporter-gold| `Mateusz Kwiatkowski (runhyve.app) <https://runhyve.app>`_
 * |supporter-std| `Gaelan D'costa <https://www.robot-disco.net>`_
 * |supporter-std| `Tenzin Lhakhang <https://confluence.tenzin.io>`_

docs/tutorial.rst

@@ -47,7 +47,7 @@ We can model this situation as two jobs:
 Install zrepl
 -------------
 
-Follow the :ref:`OS-specific installation instructions <installation>` and come back here.
+Follow the :ref:`OS-specific installation instructions <installation_toc>` and come back here.
 
 Generate TLS Certificates
 -------------------------

docs/usage.rst

@@ -38,8 +38,8 @@ CLI Overview
     * - ``zrepl migrate``
       - | perform on-disk state / ZFS property migrations
         | (see :ref:`changelog <changelog>` for details)
-    * - ``zrepl holds``
-      - list and remove holds and step bookmarks created by zrepl (see :ref:`overview <replication-cursor-and-last-received-hold>` )
+    * - ``zrepl zfs-abstractions``
+      - list and remove zrepl's abstractions on top of ZFS, e.g. holds and step bookmarks (see :ref:`overview <replication-cursor-and-last-received-hold>` )
 
 .. _usage-zrepl-daemon:
 

endpoint/context.go

@@ -3,25 +3,18 @@ package endpoint
 import (
 	"context"
 
+	"github.com/zrepl/zrepl/daemon/logging"
 	"github.com/zrepl/zrepl/logger"
 )
 
 type contextKey int
 
 const (
-	contextKeyLogger contextKey = iota
-	ClientIdentityKey
+	ClientIdentityKey contextKey = iota
 )
 
 type Logger = logger.Logger
 
-func WithLogger(ctx context.Context, log Logger) context.Context {
-	return context.WithValue(ctx, contextKeyLogger, log)
-}
-
 func getLogger(ctx context.Context) Logger {
-	if l, ok := ctx.Value(contextKeyLogger).(Logger); ok {
-		return l
-	}
-	return logger.NewNullLogger()
+	return logging.GetLogger(ctx, logging.SubsysEndpoint)
 }

endpoint/endpoint.go

@@ -2,14 +2,18 @@
 package endpoint
 
 import (
+	"bytes"
 	"context"
 	"fmt"
+	"io"
 	"path"
-	"sync"
 
+	"github.com/kr/pretty"
 	"github.com/pkg/errors"
+	"github.com/zrepl/zrepl/daemon/logging/trace"
 
 	"github.com/zrepl/zrepl/replication/logic/pdu"
+	"github.com/zrepl/zrepl/util/chainedio"
 	"github.com/zrepl/zrepl/util/chainlock"
 	"github.com/zrepl/zrepl/util/envconst"
 	"github.com/zrepl/zrepl/util/semaphore"
@@ -70,6 +74,8 @@ func (s *Sender) filterCheckFS(fs string) (*zfs.DatasetPath, error) {
 }
 
 func (s *Sender) ListFilesystems(ctx context.Context, r *pdu.ListFilesystemReq) (*pdu.ListFilesystemRes, error) {
+	defer trace.WithSpanFromStackUpdateCtx(&ctx)()
+
 	fss, err := zfs.ZFSListMapping(ctx, s.FSFilter)
 	if err != nil {
 		return nil, err
@@ -92,11 +98,13 @@ func (s *Sender) ListFilesystems(ctx context.Context, r *pdu.ListFilesystemReq)
 }
 
 func (s *Sender) ListFilesystemVersions(ctx context.Context, r *pdu.ListFilesystemVersionsReq) (*pdu.ListFilesystemVersionsRes, error) {
+	defer trace.WithSpanFromStackUpdateCtx(&ctx)()
+
 	lp, err := s.filterCheckFS(r.GetFilesystem())
 	if err != nil {
 		return nil, err
 	}
-	fsvs, err := zfs.ZFSListFilesystemVersions(lp, zfs.ListFilesystemVersionsOptions{})
+	fsvs, err := zfs.ZFSListFilesystemVersions(ctx, lp, zfs.ListFilesystemVersionsOptions{})
 	if err != nil {
 		return nil, err
 	}
@@ -104,123 +112,20 @@ func (s *Sender) ListFilesystemVersions(ctx context.Context, r *pdu.ListFilesyst
 	for i := range fsvs {
 		rfsvs[i] = pdu.FilesystemVersionFromZFS(&fsvs[i])
 	}
+
+	sendAbstractions := sendAbstractionsCacheSingleton.GetByFS(ctx, lp.ToString())
+	rSabsInfo := make([]*pdu.SendAbstraction, len(sendAbstractions))
+	for i := range rSabsInfo {
+		rSabsInfo[i] = SendAbstractionToPDU(sendAbstractions[i])
+	}
+
 	res := &pdu.ListFilesystemVersionsRes{Versions: rfsvs}
 	return res, nil
 
 }
 
-func (p *Sender) HintMostRecentCommonAncestor(ctx context.Context, r *pdu.HintMostRecentCommonAncestorReq) (*pdu.HintMostRecentCommonAncestorRes, error) {
-
-	fsp, err := p.filterCheckFS(r.GetFilesystem())
-	if err != nil {
-		return nil, err
-	}
-	fs := fsp.ToString()
-
-	log := getLogger(ctx).WithField("fs", fs).WithField("hinted_most_recent", fmt.Sprintf("%#v", r.GetSenderVersion()))
-
-	log.WithField("full_hint", r).Debug("full hint")
-
-	if r.GetSenderVersion() == nil {
-		// no common ancestor found, likely due to failed prior replication attempt
-		// => release stale step holds to prevent them from accumulating
-		//    (they can accumulate on initial replication because each inital replication step might hold a different `to`)
-		// => replication cursors cannot accumulate because we always _move_ the replication cursor
-		log.Debug("releasing all step holds on the filesystem")
-		TryReleaseStepStaleFS(ctx, fs, p.jobId)
-		return &pdu.HintMostRecentCommonAncestorRes{}, nil
-	}
-
-	// we were hinted a specific common ancestor
-
-	mostRecentVersion, err := sendArgsFromPDUAndValidateExistsAndGetVersion(ctx, fs, r.GetSenderVersion())
-	if err != nil {
-		msg := "HintMostRecentCommonAncestor rpc with nonexistent most recent version"
-		log.Warn(msg)
-		return nil, errors.Wrap(err, msg)
-	}
-
-	// move replication cursor to this position
-	destroyedCursors, err := MoveReplicationCursor(ctx, fs, mostRecentVersion, p.jobId)
-	if err == zfs.ErrBookmarkCloningNotSupported {
-		log.Debug("not creating replication cursor from bookmark because ZFS does not support it")
-		// fallthrough
-	} else if err != nil {
-		return nil, errors.Wrap(err, "cannot set replication cursor to hinted version")
-	}
-
-	// take care of stale step holds
-	log.WithField("step-holds-cleanup-mode", senderHintMostRecentCommonAncestorStepCleanupMode).
-		Debug("taking care of possibly stale step holds")
-	doStepCleanup := false
-	var stepCleanupSince *CreateTXGRangeBound
-	switch senderHintMostRecentCommonAncestorStepCleanupMode {
-	case StepCleanupNoCleanup:
-		doStepCleanup = false
-	case StepCleanupRangeSinceUnbounded:
-		doStepCleanup = true
-		stepCleanupSince = nil
-	case StepCleanupRangeSinceReplicationCursor:
-		doStepCleanup = true
-		// Use the destroyed replication cursors as indicator how far the previous replication got.
-		// To be precise: We limit the amount of visisted snapshots to exactly those snapshots
-		// created since the last successful replication cursor movement (i.e. last successful replication step)
-		//
-		// If we crash now, we'll leak the step we are about to release, but the performance gain
-		// of limiting the amount of snapshots we visit makes up for that.
-		// Users have the `zrepl holds release-stale` command to cleanup leaked step holds.
-		for _, destroyed := range destroyedCursors {
-			if stepCleanupSince == nil {
-				stepCleanupSince = &CreateTXGRangeBound{
-					CreateTXG: destroyed.GetCreateTXG(),
-					Inclusive: &zfs.NilBool{B: true},
-				}
-			} else if destroyed.GetCreateTXG() < stepCleanupSince.CreateTXG {
-				stepCleanupSince.CreateTXG = destroyed.GetCreateTXG()
-			}
-		}
-	default:
-		panic(senderHintMostRecentCommonAncestorStepCleanupMode)
-	}
-	if !doStepCleanup {
-		log.Info("skipping cleanup of prior invocations' step holds due to environment variable setting")
-	} else {
-		if err := ReleaseStepCummulativeInclusive(ctx, fs, stepCleanupSince, mostRecentVersion, p.jobId); err != nil {
-			return nil, errors.Wrap(err, "cannot cleanup prior invocation's step holds and bookmarks")
-		} else {
-			log.Info("step hold cleanup done")
-		}
-	}
-
-	return &pdu.HintMostRecentCommonAncestorRes{}, nil
-}
-
-type HintMostRecentCommonAncestorStepCleanupMode struct{ string }
-
-var (
-	StepCleanupRangeSinceReplicationCursor = HintMostRecentCommonAncestorStepCleanupMode{"range-since-replication-cursor"}
-	StepCleanupRangeSinceUnbounded         = HintMostRecentCommonAncestorStepCleanupMode{"range-since-unbounded"}
-	StepCleanupNoCleanup                   = HintMostRecentCommonAncestorStepCleanupMode{"no-cleanup"}
-)
-
-func (m HintMostRecentCommonAncestorStepCleanupMode) String() string { return string(m.string) }
-func (m *HintMostRecentCommonAncestorStepCleanupMode) Set(s string) error {
-	switch s {
-	case StepCleanupRangeSinceReplicationCursor.String():
-		*m = StepCleanupRangeSinceReplicationCursor
-	case StepCleanupRangeSinceUnbounded.String():
-		*m = StepCleanupRangeSinceUnbounded
-	case StepCleanupNoCleanup.String():
-		*m = StepCleanupNoCleanup
-	default:
-		return fmt.Errorf("unknown step cleanup mode %q", s)
-	}
-	return nil
-}
-
-var senderHintMostRecentCommonAncestorStepCleanupMode = *envconst.Var("ZREPL_ENDPOINT_SENDER_HINT_MOST_RECENT_STEP_HOLD_CLEANUP_MODE", &StepCleanupRangeSinceReplicationCursor).(*HintMostRecentCommonAncestorStepCleanupMode)
-
-var maxConcurrentZFSSendSemaphore = semaphore.New(envconst.Int64("ZREPL_ENDPOINT_MAX_CONCURRENT_SEND", 10))
+var maxConcurrentZFSSend = envconst.Int64("ZREPL_ENDPOINT_MAX_CONCURRENT_SEND", 10)
+var maxConcurrentZFSSendSemaphore = semaphore.New(maxConcurrentZFSSend)
 
 func uncheckedSendArgsFromPDU(fsv *pdu.FilesystemVersion) *zfs.ZFSSendArgVersion {
 	if fsv == nil {
@@ -241,7 +146,8 @@ func sendArgsFromPDUAndValidateExistsAndGetVersion(ctx context.Context, fs strin
 	return version, nil
 }
 
-func (s *Sender) Send(ctx context.Context, r *pdu.SendReq) (*pdu.SendRes, zfs.StreamCopier, error) {
+func (s *Sender) Send(ctx context.Context, r *pdu.SendReq) (*pdu.SendRes, io.ReadCloser, error) {
+	defer trace.WithSpanFromStackUpdateCtx(&ctx)()
 
 	_, err := s.filterCheckFS(r.Filesystem)
 	if err != nil {
@@ -309,10 +215,11 @@ func (s *Sender) Send(ctx context.Context, r *pdu.SendReq) (*pdu.SendRes, zfs.St
 		return res, nil, nil
 	}
 
-	// update replication cursor
+	// create a replication cursor for `From` (usually an idempotent no-op because SendCompleted already created it before)
+	var fromReplicationCursor Abstraction
 	if sendArgs.From != nil {
 		// For all but the first replication, this should always be a no-op because SendCompleted already moved the cursor
-		_, err = MoveReplicationCursor(ctx, sendArgs.FS, sendArgs.FromVersion, s.jobId)
+		fromReplicationCursor, err = CreateReplicationCursor(ctx, sendArgs.FS, *sendArgs.FromVersion, s.jobId) // no shadow
 		if err == zfs.ErrBookmarkCloningNotSupported {
 			getLogger(ctx).Debug("not creating replication cursor from bookmark because ZFS does not support it")
 			// fallthrough
@@ -321,9 +228,10 @@ func (s *Sender) Send(ctx context.Context, r *pdu.SendReq) (*pdu.SendRes, zfs.St
 		}
 	}
 
+	var fromHold, toHold Abstraction
 	// make sure `From` doesn't go away in order to make this step resumable
 	if sendArgs.From != nil {
-		_, err := HoldStep(ctx, sendArgs.FS, *sendArgs.FromVersion, s.jobId)
+		fromHold, err = HoldStep(ctx, sendArgs.FS, *sendArgs.FromVersion, s.jobId) // no shadow
 		if err == zfs.ErrBookmarkCloningNotSupported {
 			getLogger(ctx).Debug("not creating step bookmark because ZFS does not support it")
 			// fallthrough
@@ -332,21 +240,76 @@ func (s *Sender) Send(ctx context.Context, r *pdu.SendReq) (*pdu.SendRes, zfs.St
 		}
 	}
 	// make sure `To` doesn't go away in order to make this step resumable
-	_, err = HoldStep(ctx, sendArgs.FS, sendArgs.ToVersion, s.jobId)
+	toHold, err = HoldStep(ctx, sendArgs.FS, sendArgs.ToVersion, s.jobId)
 	if err != nil {
 		return nil, nil, errors.Wrapf(err, "cannot hold `to` version %q before starting send", sendArgs.ToVersion)
 	}
 
-	// step holds & replication cursor released / moved forward in s.SendCompleted => s.moveCursorAndReleaseSendHolds
+	// cleanup the mess that _this function_ might have created in prior failed attempts:
+	//
+	// In summary, we delete every endpoint ZFS abstraction created on this filesystem for this job id,
+	// except for the ones we just created above.
+	//
+	// This is the most robust approach to avoid leaking (= forgetting to clean up) endpoint ZFS abstractions,
+	// all under the assumption that there will only ever be one send for a (jobId,fs) combination at any given time.
+	//
+	// Note that the SendCompleted rpc can't be relied upon for this purpose:
+	// - it might be lost due to network errors,
+	// - or never be sent by a potentially malicious or buggy client,
+	// - or never be send because the replication step failed at some point
+	//   (potentially leaving a resumable state on the receiver, which is the case where we really do not want to blow away the step holds too soon.)
+	//
+	// Note further that a resuming send, due to the idempotent nature of func CreateReplicationCursor and HoldStep,
+	// will never lose its step holds because we just (idempotently re-)created them above, before attempting the cleanup.
+	func() {
+		ctx, endSpan := trace.WithSpan(ctx, "cleanup-stale-abstractions")
+		defer endSpan()
 
-	streamCopier, err := zfs.ZFSSend(ctx, sendArgs)
+		liveAbs := []Abstraction{fromHold, toHold, fromReplicationCursor}
+		keep := func(a Abstraction) (keep bool) {
+			keep = false
+			for _, k := range liveAbs {
+				keep = keep || AbstractionEquals(a, k)
+			}
+			return keep
+		}
+		check := func(obsoleteAbs []Abstraction) {
+			// last line of defense: check that we don't destroy the incremental `from` and `to`
+			// if we did that, we might be about to blow away the last common filesystem version between sender and receiver
+			mustLiveVersions := []zfs.FilesystemVersion{sendArgs.ToVersion}
+			if sendArgs.FromVersion != nil {
+				mustLiveVersions = append(mustLiveVersions, *sendArgs.FromVersion)
+			}
+			for _, staleVersion := range obsoleteAbs {
+				for _, mustLiveVersion := range mustLiveVersions {
+					if zfs.FilesystemVersionEqualIdentity(mustLiveVersion, staleVersion.GetFilesystemVersion()) {
+						panic(fmt.Sprintf("impl error: %q would be destroyed because it is considered stale but it is part of of sendArgs=%s", mustLiveVersion.String(), pretty.Sprint(sendArgs)))
+					}
+				}
+			}
+		}
+		sendAbstractionsCacheSingleton.TryBatchDestroy(ctx, s.jobId, sendArgs.FS, keep, check)
+	}()
+
+	if fromHold != nil {
+		sendAbstractionsCacheSingleton.Put(fromHold)
+	}
+	sendAbstractionsCacheSingleton.Put(toHold)
+	if fromReplicationCursor != nil {
+		sendAbstractionsCacheSingleton.Put(fromReplicationCursor)
+	}
+
+	sendStream, err := zfs.ZFSSend(ctx, sendArgs)
 	if err != nil {
+		// it's ok to not destroy the abstractions we just created here, a new send attempt will take care of it
 		return nil, nil, errors.Wrap(err, "zfs send failed")
 	}
-	return res, streamCopier, nil
+
+	return res, sendStream, nil
 }
 
 func (p *Sender) SendCompleted(ctx context.Context, r *pdu.SendCompletedReq) (*pdu.SendCompletedRes, error) {
+	defer trace.WithSpanFromStackUpdateCtx(&ctx)()
 
 	orig := r.GetOriginalReq() // may be nil, always use proto getters
 	fsp, err := p.filterCheckFS(orig.GetFilesystem())
@@ -368,76 +331,43 @@ func (p *Sender) SendCompleted(ctx context.Context, r *pdu.SendCompletedReq) (*p
 		return nil, errors.Wrap(err, "validate `to` exists")
 	}
 
-	log := getLogger(ctx).WithField("to_guid", to.Guid).
-		WithField("fs", fs).
-		WithField("to", to.RelName)
-	if from != nil {
-		log = log.WithField("from", from.RelName).WithField("from_guid", from.Guid)
+	log := func(ctx context.Context) Logger {
+		log := getLogger(ctx).WithField("to_guid", to.Guid).
+			WithField("fs", fs).
+			WithField("to", to.RelName)
+		if from != nil {
+			log = log.WithField("from", from.RelName).WithField("from_guid", from.Guid)
+		}
+		return log
 	}
 
-	log.Debug("move replication cursor to most recent common version")
-	destroyedCursors, err := MoveReplicationCursor(ctx, fs, to, p.jobId)
+	toReplicationCursor, err := CreateReplicationCursor(ctx, fs, to, p.jobId)
 	if err != nil {
 		if err == zfs.ErrBookmarkCloningNotSupported {
-			log.Debug("not setting replication cursor, bookmark cloning not supported")
+			log(ctx).Debug("not setting replication cursor, bookmark cloning not supported")
 		} else {
 			msg := "cannot move replication cursor, keeping hold on `to` until successful"
-			log.WithError(err).Error(msg)
+			log(ctx).WithError(err).Error(msg)
 			err = errors.Wrap(err, msg)
-			// it is correct to not release the hold if we can't move the cursor!
+			// it is correct to not destroy from and to step holds if we can't move the cursor!
 			return &pdu.SendCompletedRes{}, err
 		}
 	} else {
-		log.Info("successfully moved replication cursor")
+		log(ctx).WithField("to_cursor", toReplicationCursor.String()).Info("successfully created `to` replication cursor")
 	}
 
-	// kick off releasing of step holds / bookmarks
-	// if we fail to release them, don't bother the caller:
-	// they are merely an implementation detail on the sender for better resumability
-	var wg sync.WaitGroup
-	wg.Add(2)
-	go func() {
-		defer wg.Done()
-		log.Debug("release step-hold of or step-bookmark on `to`")
-		err = ReleaseStep(ctx, fs, to, p.jobId)
-		if err != nil {
-			log.WithError(err).Error("cannot release step-holds on or destroy step-bookmark of `to`")
-		} else {
-			log.Info("successfully released step-holds on or destroyed step-bookmark of `to`")
-		}
-
-	}()
-	go func() {
-		defer wg.Done()
-		if from == nil {
-			return
-		}
-		log.Debug("release step-hold of or step-bookmark on `from`")
-		err := ReleaseStep(ctx, fs, *from, p.jobId)
-		if err != nil {
-			if dne, ok := err.(*zfs.DatasetDoesNotExist); ok {
-				// If bookmark cloning is not supported, `from` might be the old replication cursor
-				// and thus have already been destroyed by MoveReplicationCursor above
-				// In that case, nonexistence of `from` is not an error, otherwise it is.
-				for _, c := range destroyedCursors {
-					if c.GetFullPath() == dne.Path {
-						log.Info("`from` was a replication cursor and has already been destroyed")
-						return
-					}
-				}
-				// fallthrough
-			}
-			log.WithError(err).Error("cannot release step-holds on or destroy step-bookmark of `from`")
-		} else {
-			log.Info("successfully released step-holds on or destroyed step-bookmark of `from`")
-		}
-	}()
-	wg.Wait()
+	keep := func(a Abstraction) bool {
+		return AbstractionEquals(a, toReplicationCursor)
+	}
+	sendAbstractionsCacheSingleton.TryBatchDestroy(ctx, p.jobId, fs, keep, nil)
 
 	return &pdu.SendCompletedRes{}, nil
+
 }
 
 func (p *Sender) DestroySnapshots(ctx context.Context, req *pdu.DestroySnapshotsReq) (*pdu.DestroySnapshotsRes, error) {
+	defer trace.WithSpanFromStackUpdateCtx(&ctx)()
+
 	dp, err := p.filterCheckFS(req.Filesystem)
 	if err != nil {
 		return nil, err
@@ -446,6 +376,8 @@ func (p *Sender) DestroySnapshots(ctx context.Context, req *pdu.DestroySnapshots
 }
 
 func (p *Sender) Ping(ctx context.Context, req *pdu.PingReq) (*pdu.PingRes, error) {
+	defer trace.WithSpanFromStackUpdateCtx(&ctx)()
+
 	res := pdu.PingRes{
 		Echo: req.GetMessage(),
 	}
@@ -453,14 +385,20 @@ func (p *Sender) Ping(ctx context.Context, req *pdu.PingReq) (*pdu.PingRes, erro
 }
 
 func (p *Sender) PingDataconn(ctx context.Context, req *pdu.PingReq) (*pdu.PingRes, error) {
+	defer trace.WithSpanFromStackUpdateCtx(&ctx)()
+
 	return p.Ping(ctx, req)
 }
 
 func (p *Sender) WaitForConnectivity(ctx context.Context) error {
+	defer trace.WithSpanFromStackUpdateCtx(&ctx)()
+
 	return nil
 }
 
 func (p *Sender) ReplicationCursor(ctx context.Context, req *pdu.ReplicationCursorReq) (*pdu.ReplicationCursorRes, error) {
+	defer trace.WithSpanFromStackUpdateCtx(&ctx)()
+
 	dp, err := p.filterCheckFS(req.Filesystem)
 	if err != nil {
 		return nil, err
@@ -476,7 +414,7 @@ func (p *Sender) ReplicationCursor(ctx context.Context, req *pdu.ReplicationCurs
 	return &pdu.ReplicationCursorRes{Result: &pdu.ReplicationCursorRes_Guid{Guid: cursor.Guid}}, nil
 }
 
-func (p *Sender) Receive(ctx context.Context, r *pdu.ReceiveReq, receive zfs.StreamCopier) (*pdu.ReceiveRes, error) {
+func (p *Sender) Receive(ctx context.Context, r *pdu.ReceiveReq, _ io.ReadCloser) (*pdu.ReceiveRes, error) {
 	return nil, fmt.Errorf("sender does not implement Receive()")
 }
 
@@ -591,6 +529,15 @@ func (f subroot) MapToLocal(fs string) (*zfs.DatasetPath, error) {
 }
 
 func (s *Receiver) ListFilesystems(ctx context.Context, req *pdu.ListFilesystemReq) (*pdu.ListFilesystemRes, error) {
+	defer trace.WithSpanFromStackUpdateCtx(&ctx)()
+
+	// first make sure that root_fs is imported
+	if rphs, err := zfs.ZFSGetFilesystemPlaceholderState(ctx, s.conf.RootWithoutClientComponent); err != nil {
+		return nil, errors.Wrap(err, "cannot determine whether root_fs exists")
+	} else if !rphs.FSExists {
+		return nil, errors.New("root_fs does not exist")
+	}
+
 	root := s.clientRootFromCtx(ctx)
 	filtered, err := zfs.ZFSListMapping(ctx, subroot{root})
 	if err != nil {
@@ -641,6 +588,8 @@ func (s *Receiver) ListFilesystems(ctx context.Context, req *pdu.ListFilesystemR
 }
 
 func (s *Receiver) ListFilesystemVersions(ctx context.Context, req *pdu.ListFilesystemVersionsReq) (*pdu.ListFilesystemVersionsRes, error) {
+	defer trace.WithSpanFromStackUpdateCtx(&ctx)()
+
 	root := s.clientRootFromCtx(ctx)
 	lp, err := subroot{root}.MapToLocal(req.GetFilesystem())
 	if err != nil {
@@ -648,7 +597,7 @@ func (s *Receiver) ListFilesystemVersions(ctx context.Context, req *pdu.ListFile
 	}
 	// TODO share following code with sender
 
-	fsvs, err := zfs.ZFSListFilesystemVersions(lp, zfs.ListFilesystemVersionsOptions{})
+	fsvs, err := zfs.ZFSListFilesystemVersions(ctx, lp, zfs.ListFilesystemVersionsOptions{})
 	if err != nil {
 		return nil, err
 	}
@@ -662,6 +611,8 @@ func (s *Receiver) ListFilesystemVersions(ctx context.Context, req *pdu.ListFile
 }
 
 func (s *Receiver) Ping(ctx context.Context, req *pdu.PingReq) (*pdu.PingRes, error) {
+	defer trace.WithSpanFromStackUpdateCtx(&ctx)()
+
 	res := pdu.PingRes{
 		Echo: req.GetMessage(),
 	}
@@ -669,24 +620,30 @@ func (s *Receiver) Ping(ctx context.Context, req *pdu.PingReq) (*pdu.PingRes, er
 }
 
 func (s *Receiver) PingDataconn(ctx context.Context, req *pdu.PingReq) (*pdu.PingRes, error) {
+	defer trace.WithSpanFromStackUpdateCtx(&ctx)()
 	return s.Ping(ctx, req)
 }
 
 func (s *Receiver) WaitForConnectivity(ctx context.Context) error {
+	defer trace.WithSpanFromStackUpdateCtx(&ctx)()
 	return nil
 }
 
-func (s *Receiver) ReplicationCursor(context.Context, *pdu.ReplicationCursorReq) (*pdu.ReplicationCursorRes, error) {
+func (s *Receiver) ReplicationCursor(ctx context.Context, _ *pdu.ReplicationCursorReq) (*pdu.ReplicationCursorRes, error) {
+	defer trace.WithSpanFromStackUpdateCtx(&ctx)()
 	return nil, fmt.Errorf("ReplicationCursor not implemented for Receiver")
 }
 
-func (s *Receiver) Send(ctx context.Context, req *pdu.SendReq) (*pdu.SendRes, zfs.StreamCopier, error) {
+func (s *Receiver) Send(ctx context.Context, req *pdu.SendReq) (*pdu.SendRes, io.ReadCloser, error) {
+	defer trace.WithSpanFromStackUpdateCtx(&ctx)()
 	return nil, nil, fmt.Errorf("receiver does not implement Send()")
 }
 
 var maxConcurrentZFSRecvSemaphore = semaphore.New(envconst.Int64("ZREPL_ENDPOINT_MAX_CONCURRENT_RECV", 10))
 
-func (s *Receiver) Receive(ctx context.Context, req *pdu.ReceiveReq, receive zfs.StreamCopier) (*pdu.ReceiveRes, error) {
+func (s *Receiver) Receive(ctx context.Context, req *pdu.ReceiveReq, receive io.ReadCloser) (*pdu.ReceiveRes, error) {
+	defer trace.WithSpanFromStackUpdateCtx(&ctx)()
+
 	getLogger(ctx).Debug("incoming Receive")
 	defer receive.Close()
 
@@ -765,21 +722,30 @@ func (s *Receiver) Receive(ctx context.Context, req *pdu.ReceiveReq, receive zfs
 		return nil, visitErr
 	}
 
+	log := getLogger(ctx).WithField("proto_fs", req.GetFilesystem()).WithField("local_fs", lp.ToString())
+
 	// determine whether we need to rollback the filesystem / change its placeholder state
 	var clearPlaceholderProperty bool
 	var recvOpts zfs.RecvOptions
 	ph, err := zfs.ZFSGetFilesystemPlaceholderState(ctx, lp)
-	if err == nil && ph.FSExists && ph.IsPlaceholder {
+	if err != nil {
+		return nil, errors.Wrap(err, "cannot get placeholder state")
+	}
+	log.WithField("placeholder_state", fmt.Sprintf("%#v", ph)).Debug("placeholder state")
+	if ph.FSExists && ph.IsPlaceholder {
 		recvOpts.RollbackAndForceRecv = true
 		clearPlaceholderProperty = true
 	}
+
 	if clearPlaceholderProperty {
+		log.Info("clearing placeholder property")
 		if err := zfs.ZFSSetPlaceholder(ctx, lp, false); err != nil {
 			return nil, fmt.Errorf("cannot clear placeholder property for forced receive: %s", err)
 		}
 	}
 
 	if req.ClearResumeToken && ph.FSExists {
+		log.Info("clearing resume token")
 		if err := zfs.ZFSRecvClearResumeToken(ctx, lp.ToString()); err != nil {
 			return nil, errors.Wrap(err, "cannot clear resume token")
 		}
@@ -790,7 +756,7 @@ func (s *Receiver) Receive(ctx context.Context, req *pdu.ReceiveReq, receive zfs
 		return nil, errors.Wrap(err, "cannot determine whether we can use resumable send & recv")
 	}
 
-	getLogger(ctx).Debug("acquire concurrent recv semaphore")
+	log.Debug("acquire concurrent recv semaphore")
 	// TODO use try-acquire and fail with resource-exhaustion rpc status
 	// => would require handling on the client-side
 	// => this is a dataconn endpoint, doesn't have the status code semantics of gRPC
@@ -800,14 +766,89 @@ func (s *Receiver) Receive(ctx context.Context, req *pdu.ReceiveReq, receive zfs
 	}
 	defer guard.Release()
 
-	getLogger(ctx).WithField("opts", fmt.Sprintf("%#v", recvOpts)).Debug("start receive command")
+	var peek bytes.Buffer
+	var MaxPeek = envconst.Int64("ZREPL_ENDPOINT_RECV_PEEK_SIZE", 1<<20)
+	log.WithField("max_peek_bytes", MaxPeek).Info("peeking incoming stream")
+	if _, err := io.Copy(&peek, io.LimitReader(receive, MaxPeek)); err != nil {
+		log.WithError(err).Error("cannot read peek-buffer from send stream")
+	}
+	var peekCopy bytes.Buffer
+	if n, err := peekCopy.Write(peek.Bytes()); err != nil || n != peek.Len() {
+		panic(peek.Len())
+	}
+
+	log.WithField("opts", fmt.Sprintf("%#v", recvOpts)).Debug("start receive command")
 
 	snapFullPath := to.FullPath(lp.ToString())
-	if err := zfs.ZFSRecv(ctx, lp.ToString(), to, receive, recvOpts); err != nil {
-		getLogger(ctx).
+	if err := zfs.ZFSRecv(ctx, lp.ToString(), to, chainedio.NewChainedReader(&peek, receive), recvOpts); err != nil {
+
+		// best-effort rollback of placeholder state if the recv didn't start
+		_, resumableStatePresent := err.(*zfs.RecvFailedWithResumeTokenErr)
+		disablePlaceholderRestoration := envconst.Bool("ZREPL_ENDPOINT_DISABLE_PLACEHOLDER_RESTORATION", false)
+		placeholderRestored := !ph.IsPlaceholder
+		if !disablePlaceholderRestoration && !resumableStatePresent && recvOpts.RollbackAndForceRecv && ph.FSExists && ph.IsPlaceholder && clearPlaceholderProperty {
+			log.Info("restoring placeholder property")
+			if phErr := zfs.ZFSSetPlaceholder(ctx, lp, true); phErr != nil {
+				log.WithError(phErr).Error("cannot restore placeholder property after failed receive, subsequent replications will likely fail with a different error")
+				// fallthrough
+			} else {
+				placeholderRestored = true
+			}
+			// fallthrough
+		}
+
+		// deal with failing initial encrypted send & recv
+		if _, ok := err.(*zfs.RecvDestroyOrOverwriteEncryptedErr); ok && ph.IsPlaceholder && placeholderRestored {
+			msg := `cannot automatically replace placeholder filesystem with incoming send stream - please see receive-side log for details`
+			err := errors.New(msg)
+			log.Error(msg)
+
+			log.Error(`zrepl creates placeholder filesystems on the receiving side of a replication to match the sending side's dataset hierarchy`)
+			log.Error(`zrepl uses zfs receive -F to replace those placeholders with incoming full sends`)
+			log.Error(`OpenZFS native encryption prohibits zfs receive -F for encrypted filesystems`)
+			log.Error(`the current zrepl placeholder filesystem concept is thus incompatible with OpenZFS native encryption`)
+
+			tempStartFullRecvFS := lp.Copy().ToString() + ".zrepl.initial-recv"
+			tempStartFullRecvFSDP, dpErr := zfs.NewDatasetPath(tempStartFullRecvFS)
+			if dpErr != nil {
+				log.WithError(dpErr).Error("cannot determine temporary filesystem name for initial encrypted recv workaround")
+				return nil, err // yes, err, not dpErr
+			}
+
+			log := log.WithField("temp_recv_fs", tempStartFullRecvFS)
+			log.Error(`as a workaround, zrepl will now attempt to re-receive the beginning of the stream into a temporary filesystem temp_recv_fs`)
+			log.Error(`if that step succeeds: shut down zrepl and use 'zfs rename' to swap temp_recv_fs with local_fs, then restart zrepl`)
+			log.Error(`replication will then resume using resumable send+recv`)
+
+			tempPH, phErr := zfs.ZFSGetFilesystemPlaceholderState(ctx, tempStartFullRecvFSDP)
+			if phErr != nil {
+				log.WithError(phErr).Error("cannot determine placeholder state of temp_recv_fs")
+				return nil, err // yes, err, not dpErr
+			}
+			if tempPH.FSExists {
+				log.Error("temp_recv_fs already exists, assuming a (partial) initial recv to that filesystem has already been done")
+				return nil, err
+			}
+
+			recvOpts.RollbackAndForceRecv = false
+			recvOpts.SavePartialRecvState = true
+			rerecvErr := zfs.ZFSRecv(ctx, tempStartFullRecvFS, to, chainedio.NewChainedReader(&peekCopy), recvOpts)
+			if _, isResumable := rerecvErr.(*zfs.RecvFailedWithResumeTokenErr); rerecvErr == nil || isResumable {
+				log.Error("completed re-receive into temporary filesystem temp_recv_fs, now shut down zrepl and use zfs rename to swap temp_recv_fs with local_fs")
+			} else {
+				log.WithError(rerecvErr).Error("failed to receive the beginning of the stream into temporary filesystem temp_recv_fs")
+				log.Error("we advise you to collect the error log and current configuration, open an issue on GitHub, and revert to your previous configuration in the meantime")
+			}
+
+			log.Error(`if you would like to see improvements to this situation, please open an issue on GitHub`)
+			return nil, err
+		}
+
+		log.
 			WithError(err).
 			WithField("opts", fmt.Sprintf("%#v", recvOpts)).
 			Error("zfs receive failed")
+
 		return nil, err
 	}
 
@@ -815,13 +856,13 @@ func (s *Receiver) Receive(ctx context.Context, req *pdu.ReceiveReq, receive zfs
 	toRecvd, err := to.ValidateExistsAndGetVersion(ctx, lp.ToString())
 	if err != nil {
 		msg := "receive request's `To` version does not match what we received in the stream"
-		getLogger(ctx).WithError(err).WithField("snap", snapFullPath).Error(msg)
-		getLogger(ctx).Error("aborting recv request, but keeping received snapshot for inspection")
+		log.WithError(err).WithField("snap", snapFullPath).Error(msg)
+		log.Error("aborting recv request, but keeping received snapshot for inspection")
 		return nil, errors.Wrap(err, msg)
 	}
 
 	if s.conf.UpdateLastReceivedHold {
-		getLogger(ctx).Debug("move last-received-hold")
+		log.Debug("move last-received-hold")
 		if err := MoveLastReceivedHold(ctx, lp.ToString(), toRecvd, s.conf.JobID); err != nil {
 			return nil, errors.Wrap(err, "cannot move last-received-hold")
 		}
@@ -831,6 +872,8 @@ func (s *Receiver) Receive(ctx context.Context, req *pdu.ReceiveReq, receive zfs
 }
 
 func (s *Receiver) DestroySnapshots(ctx context.Context, req *pdu.DestroySnapshotsReq) (*pdu.DestroySnapshotsRes, error) {
+	defer trace.WithSpanFromStackUpdateCtx(&ctx)()
+
 	root := s.clientRootFromCtx(ctx)
 	lp, err := subroot{root}.MapToLocal(req.Filesystem)
 	if err != nil {
@@ -839,16 +882,9 @@ func (s *Receiver) DestroySnapshots(ctx context.Context, req *pdu.DestroySnapsho
 	return doDestroySnapshots(ctx, lp, req.Snapshots)
 }
 
-func (p *Receiver) HintMostRecentCommonAncestor(ctx context.Context, r *pdu.HintMostRecentCommonAncestorReq) (*pdu.HintMostRecentCommonAncestorRes, error) {
-	// we don't move last-received-hold as part of this hint
-	// because that wouldn't give us any benefit wrt resumability.
-	//
-	// Other reason: the replication logic that issues this RPC would require refactoring
-	// to include the receiver's FilesystemVersion in the request)
-	return &pdu.HintMostRecentCommonAncestorRes{}, nil
-}
+func (p *Receiver) SendCompleted(ctx context.Context, _ *pdu.SendCompletedReq) (*pdu.SendCompletedRes, error) {
+	defer trace.WithSpanFromStackUpdateCtx(&ctx)()
 
-func (p *Receiver) SendCompleted(context.Context, *pdu.SendCompletedReq) (*pdu.SendCompletedRes, error) {
 	return &pdu.SendCompletedRes{}, nil
 }
 
@@ -870,7 +906,7 @@ func doDestroySnapshots(ctx context.Context, lp *zfs.DatasetPath, snaps []*pdu.F
 			ErrOut:     &errs[i],
 		}
 	}
-	zfs.ZFSDestroyFilesystemVersions(reqs)
+	zfs.ZFSDestroyFilesystemVersions(ctx, reqs)
 	for i := range reqs {
 		if errs[i] != nil {
 			if de, ok := errs[i].(*zfs.DestroySnapshotsError); ok && len(de.Reason) == 1 {

endpoint/endpoint_metrics.go

@@ -0,0 +1,7 @@
+package endpoint
+
+import "github.com/prometheus/client_golang/prometheus"
+
+func RegisterMetrics(r prometheus.Registerer) {
+	r.MustRegister(sendAbstractionsCacheMetrics.count)
+}

endpoint/endpoint_send_abstractions_cache.go

@@ -0,0 +1,232 @@
+package endpoint
+
+import (
+	"context"
+	"fmt"
+	"sync"
+
+	"github.com/prometheus/client_golang/prometheus"
+	"github.com/zrepl/zrepl/daemon/logging/trace"
+	"github.com/zrepl/zrepl/util/chainlock"
+)
+
+var sendAbstractionsCacheMetrics struct {
+	count prometheus.Gauge
+}
+
+func init() {
+	sendAbstractionsCacheMetrics.count = prometheus.NewGauge(prometheus.GaugeOpts{
+		Namespace: "zrepl",
+		Subsystem: "endpoint",
+		Name:      "send_abstractions_cache_entry_count",
+		Help:      "number of send abstractions tracked in the sendAbstractionsCache data structure",
+	})
+}
+
+var sendAbstractionsCacheSingleton = newSendAbstractionsCache()
+
+func SendAbstractionsCacheInvalidate(fs string) {
+	sendAbstractionsCacheSingleton.InvalidateFSCache(fs)
+}
+
+type sendAbstractionsCacheDidLoadFSState int
+
+const (
+	sendAbstractionsCacheDidLoadFSStateNo sendAbstractionsCacheDidLoadFSState = iota // 0-value has meaning
+	sendAbstractionsCacheDidLoadFSStateInProgress
+	sendAbstractionsCacheDidLoadFSStateDone
+)
+
+type sendAbstractionsCache struct {
+	mtx              chainlock.L
+	abstractions     []Abstraction
+	didLoadFS        map[string]sendAbstractionsCacheDidLoadFSState
+	didLoadFSChanged *sync.Cond
+}
+
+func newSendAbstractionsCache() *sendAbstractionsCache {
+	c := &sendAbstractionsCache{
+		didLoadFS: make(map[string]sendAbstractionsCacheDidLoadFSState),
+	}
+	c.didLoadFSChanged = c.mtx.NewCond()
+	return c
+}
+
+func (s *sendAbstractionsCache) Put(a Abstraction) {
+	defer s.mtx.Lock().Unlock()
+
+	var zeroJobId JobID
+	if a.GetJobID() == nil {
+		panic("abstraction must not have nil job id")
+	} else if *a.GetJobID() == zeroJobId {
+		panic(fmt.Sprintf("abstraction must not have zero-value job id: %s", a))
+	}
+
+	s.abstractions = append(s.abstractions, a)
+	sendAbstractionsCacheMetrics.count.Set(float64(len(s.abstractions)))
+}
+
+func (s *sendAbstractionsCache) InvalidateFSCache(fs string) {
+	// FIXME: O(n)
+	newAbs := make([]Abstraction, 0, len(s.abstractions))
+	for _, a := range s.abstractions {
+		if a.GetFS() != fs {
+			newAbs = append(newAbs, a)
+		}
+	}
+	s.abstractions = newAbs
+	sendAbstractionsCacheMetrics.count.Set(float64(len(s.abstractions)))
+
+	s.didLoadFS[fs] = sendAbstractionsCacheDidLoadFSStateNo
+	s.didLoadFSChanged.Broadcast()
+
+}
+
+func (s *sendAbstractionsCache) GetByFS(ctx context.Context, fs string) (ret []Abstraction) {
+	defer s.mtx.Lock().Unlock()
+	defer trace.WithSpanFromStackUpdateCtx(&ctx)()
+	if fs == "" {
+		panic("must not pass zero-value fs")
+	}
+
+	s.tryLoadOnDiskSendAbstractions(ctx, fs)
+
+	for _, a := range s.abstractions {
+		if a.GetFS() == fs {
+			ret = append(ret, a)
+		}
+	}
+
+	return ret
+}
+
+// - logs errors in getting on-disk abstractions
+// - only fetches on-disk abstractions once, but every time from the in-memory store
+//
+// That means that for precise results, all abstractions created by the endpoint must be .Put into this cache.
+func (s *sendAbstractionsCache) GetAndDeleteByJobIDAndFS(ctx context.Context, jobID JobID, fs string) (ret []Abstraction) {
+	defer s.mtx.Lock().Unlock()
+	defer trace.WithSpanFromStackUpdateCtx(&ctx)()
+	var zeroJobId JobID
+	if jobID == zeroJobId {
+		panic("must not pass zero-value job id")
+	}
+	if fs == "" {
+		panic("must not pass zero-value fs")
+	}
+
+	s.tryLoadOnDiskSendAbstractions(ctx, fs)
+
+	// FIXME O(n)
+	var remaining []Abstraction
+	for _, a := range s.abstractions {
+		aJobId := *a.GetJobID()
+		aFS := a.GetFS()
+		if aJobId == jobID && aFS == fs {
+			ret = append(ret, a)
+		} else {
+			remaining = append(remaining, a)
+		}
+	}
+	s.abstractions = remaining
+	sendAbstractionsCacheMetrics.count.Set(float64(len(s.abstractions)))
+
+	return ret
+}
+
+// caller must hold s.mtx
+func (s *sendAbstractionsCache) tryLoadOnDiskSendAbstractions(ctx context.Context, fs string) {
+	for s.didLoadFS[fs] != sendAbstractionsCacheDidLoadFSStateDone {
+		if s.didLoadFS[fs] == sendAbstractionsCacheDidLoadFSStateInProgress {
+			s.didLoadFSChanged.Wait()
+			continue
+		}
+		if s.didLoadFS[fs] != sendAbstractionsCacheDidLoadFSStateNo {
+			panic(fmt.Sprintf("unreachable: %v", s.didLoadFS[fs]))
+		}
+
+		s.didLoadFS[fs] = sendAbstractionsCacheDidLoadFSStateInProgress
+		defer s.didLoadFSChanged.Broadcast()
+
+		var onDiskAbs []Abstraction
+		var err error
+		s.mtx.DropWhile(func() {
+			onDiskAbs, err = s.tryLoadOnDiskSendAbstractionsImpl(ctx, fs) // no shadow
+		})
+
+		if err != nil {
+			s.didLoadFS[fs] = sendAbstractionsCacheDidLoadFSStateNo
+			getLogger(ctx).WithField("fs", fs).WithError(err).Error("cannot list send step abstractions for filesystem")
+		} else {
+			s.didLoadFS[fs] = sendAbstractionsCacheDidLoadFSStateDone
+			s.abstractions = append(s.abstractions, onDiskAbs...)
+			getLogger(ctx).WithField("fs", fs).WithField("abstractions", onDiskAbs).Debug("loaded step abstractions for filesystem")
+		}
+		return
+	}
+}
+
+// caller should _not hold s.mtx
+func (s *sendAbstractionsCache) tryLoadOnDiskSendAbstractionsImpl(ctx context.Context, fs string) ([]Abstraction, error) {
+	defer trace.WithSpanFromStackUpdateCtx(&ctx)()
+
+	q := ListZFSHoldsAndBookmarksQuery{
+		FS: ListZFSHoldsAndBookmarksQueryFilesystemFilter{
+			FS: &fs,
+		},
+		JobID: nil,
+		What: AbstractionTypeSet{
+			AbstractionStepHold:                    true,
+			AbstractionStepBookmark:                true,
+			AbstractionReplicationCursorBookmarkV2: true,
+		},
+		Concurrency: 1,
+	}
+	abs, absErrs, err := ListAbstractions(ctx, q)
+	if err != nil {
+		return nil, err
+	}
+	// safe to ignore absErrs here, this is best-effort cleanup
+	if len(absErrs) > 0 {
+		return nil, ListAbstractionsErrors(absErrs)
+	}
+	return abs, nil
+}
+
+func (s *sendAbstractionsCache) TryBatchDestroy(ctx context.Context, jobId JobID, fs string, keep func(a Abstraction) bool, check func(willDestroy []Abstraction)) {
+	// no s.mtx, we only use the public interface in this function
+
+	defer trace.WithSpanFromStackUpdateCtx(&ctx)()
+
+	allSendStepAbstractions := s.GetAndDeleteByJobIDAndFS(ctx, jobId, fs)
+
+	var obsoleteAbs []Abstraction
+	for _, a := range allSendStepAbstractions {
+		if !keep(a) {
+			obsoleteAbs = append(obsoleteAbs, a)
+		}
+	}
+
+	if check != nil {
+		check(obsoleteAbs)
+	}
+
+	hadErr := false
+	for res := range BatchDestroy(ctx, obsoleteAbs) {
+		if res.DestroyErr != nil {
+			hadErr = true
+			getLogger(ctx).
+				WithField("abstraction", res.Abstraction).
+				WithError(res.DestroyErr).
+				Error("cannot destroy stale send step abstraction")
+		} else {
+			getLogger(ctx).
+				WithField("abstraction", res.Abstraction).
+				Info("destroyed stale send step abstraction")
+		}
+	}
+	if hadErr {
+		s.InvalidateFSCache(fs)
+	}
+
+}

endpoint/endpoint_zfs_abstraction.go

@@ -11,6 +11,7 @@ import (
 
 	"github.com/pkg/errors"
 
+	"github.com/zrepl/zrepl/daemon/logging/trace"
 	"github.com/zrepl/zrepl/util/envconst"
 	"github.com/zrepl/zrepl/util/semaphore"
 	"github.com/zrepl/zrepl/zfs"
@@ -53,6 +54,30 @@ type Abstraction interface {
 	json.Marshaler
 }
 
+func AbstractionEquals(a, b Abstraction) bool {
+	if (a != nil) != (b != nil) {
+		return false
+	}
+	if a == nil && b == nil {
+		return true
+	}
+	var aJobId, bJobId JobID
+	if aJid := a.GetJobID(); aJid != nil {
+		aJobId = *aJid
+	}
+	if bJid := b.GetJobID(); bJid != nil {
+		bJobId = *bJid
+	}
+	return a.GetType() == b.GetType() &&
+		a.GetFS() == b.GetFS() &&
+		a.GetName() == b.GetName() &&
+		a.GetFullPath() == b.GetFullPath() &&
+		aJobId == bJobId &&
+		a.GetCreateTXG() == b.GetCreateTXG() &&
+		zfs.FilesystemVersionEqualIdentity(a.GetFilesystemVersion(), b.GetFilesystemVersion()) &&
+		a.String() == b.String()
+}
+
 func (t AbstractionType) Validate() error {
 	switch t {
 	case AbstractionStepBookmark:
@@ -467,7 +492,7 @@ func (e ListAbstractionsErrors) Error() string {
 	}
 	msgs := make([]string, len(e))
 	for i := range e {
-		msgs[i] = e.Error()
+		msgs[i] = e[i].Error()
 	}
 	return fmt.Sprintf("list endpoint abstractions: multiple errors:\n%s", strings.Join(msgs, "\n"))
 }
@@ -529,15 +554,16 @@ func ListAbstractionsStreamed(ctx context.Context, query ListZFSHoldsAndBookmark
 	}
 
 	sem := semaphore.New(int64(query.Concurrency))
+	ctx, endTask := trace.WithTask(ctx, "list-abstractions-streamed-producer")
 	go func() {
+		defer endTask()
 		defer close(out)
 		defer close(outErrs)
-		var wg sync.WaitGroup
-		defer wg.Wait()
+
+		_, add, wait := trace.WithTaskGroup(ctx, "list-abstractions-impl-fs")
+		defer wait()
 		for i := range fss {
-			wg.Add(1)
-			go func(i int) {
-				defer wg.Done()
+			add(func(ctx context.Context) {
 				g, err := sem.Acquire(ctx)
 				if err != nil {
 					errCb(err, fss[i], err.Error())
@@ -547,7 +573,7 @@ func ListAbstractionsStreamed(ctx context.Context, query ListZFSHoldsAndBookmark
 					defer g.Release()
 					listAbstractionsImplFS(ctx, fss[i], &query, emitAbstraction, errCb)
 				}()
-			}(i)
+			})
 		}
 	}()
 
@@ -573,7 +599,7 @@ func listAbstractionsImplFS(ctx context.Context, fs string, query *ListZFSHoldsA
 			whatTypes[zfs.Snapshot] = true
 		}
 	}
-	fsvs, err := zfs.ZFSListFilesystemVersions(fsp, zfs.ListFilesystemVersionsOptions{
+	fsvs, err := zfs.ZFSListFilesystemVersions(ctx, fsp, zfs.ListFilesystemVersionsOptions{
 		Types: whatTypes,
 	})
 	if err != nil {
@@ -588,23 +614,23 @@ func listAbstractionsImplFS(ctx context.Context, fs string, query *ListZFSHoldsA
 			panic("implementation error: extractors misconfigured for " + at)
 		}
 		for _, v := range fsvs {
-			var a Abstraction
 			if v.Type == zfs.Bookmark && bmE != nil {
-				a = bmE(fsp, v)
+				if a := bmE(fsp, v); a != nil {
+					emitCandidate(a)
+				}
 			}
-			if v.Type == zfs.Snapshot && holdE != nil && query.CreateTXG.Contains(v.GetCreateTXG()) && (!v.UserRefs.Valid || v.UserRefs.Value > 0) {
+			if v.Type == zfs.Snapshot && holdE != nil && query.CreateTXG.Contains(v.GetCreateTXG()) && (!v.UserRefs.Valid || v.UserRefs.Value > 0) { // FIXME review v.UserRefsValid
 				holds, err := zfs.ZFSHolds(ctx, fsp.ToString(), v.Name)
 				if err != nil {
 					errCb(err, v.ToAbsPath(fsp), "get hold on snap")
 					continue
 				}
 				for _, tag := range holds {
-					a = holdE(fsp, v, tag)
+					if a := holdE(fsp, v, tag); a != nil {
+						emitCandidate(a)
+					}
 				}
 			}
-			if a != nil {
-				emitCandidate(a)
-			}
 		}
 	}
 }
@@ -731,6 +757,9 @@ func listStaleFiltering(abs []Abstraction, sinceBound *CreateTXGRangeBound) *Sta
 	}
 	stepFirstNotStaleCandidates := make(map[fsAndJobId]stepFirstNotStaleCandidate) // empty map => will always return nil
 	for _, a := range abs {
+		if a.GetJobID() == nil {
+			continue // already put those into always-live list noJobId in above loop
+		}
 		key := fsAndJobId{a.GetFS(), *a.GetJobID()}
 		c := stepFirstNotStaleCandidates[key]
 

endpoint/endpoint_zfs_abstraction_cursor_and_last_received_hold.go

@@ -118,31 +118,32 @@ func GetReplicationCursors(ctx context.Context, dp *zfs.DatasetPath, jobID JobID
 	return candidates, nil
 }
 
-type ReplicationCursorTarget interface {
-	IsSnapshot() bool
-	GetGuid() uint64
-	GetCreateTXG() uint64
-	ToSendArgVersion() zfs.ZFSSendArgVersion
-}
-
-// `target` is validated before replication cursor is set. if validation fails, the cursor is not moved.
+// idempotently create a replication cursor targeting `target`
 //
 // returns ErrBookmarkCloningNotSupported if version is a bookmark and bookmarking bookmarks is not supported by ZFS
-func MoveReplicationCursor(ctx context.Context, fs string, target ReplicationCursorTarget, jobID JobID) (destroyedCursors []Abstraction, err error) {
-
-	if !target.IsSnapshot() {
-		return nil, zfs.ErrBookmarkCloningNotSupported
-	}
+func CreateReplicationCursor(ctx context.Context, fs string, target zfs.FilesystemVersion, jobID JobID) (a Abstraction, err error) {
 
 	bookmarkname, err := ReplicationCursorBookmarkName(fs, target.GetGuid(), jobID)
 	if err != nil {
 		return nil, errors.Wrap(err, "determine replication cursor name")
 	}
 
-	// idempotently create bookmark (guid is encoded in it, hence we'll most likely add a new one
-	// cleanup the old one afterwards
+	if target.IsBookmark() && target.GetName() == bookmarkname {
+		return &bookmarkBasedAbstraction{
+			Type:              AbstractionReplicationCursorBookmarkV2,
+			FS:                fs,
+			FilesystemVersion: target,
+			JobID:             jobID,
+		}, nil
+	}
 
-	err = zfs.ZFSBookmark(ctx, fs, target.ToSendArgVersion(), bookmarkname)
+	if !target.IsSnapshot() {
+		return nil, zfs.ErrBookmarkCloningNotSupported
+	}
+
+	// idempotently create bookmark (guid is encoded in it)
+
+	cursorBookmark, err := zfs.ZFSBookmark(ctx, fs, target, bookmarkname)
 	if err != nil {
 		if err == zfs.ErrBookmarkCloningNotSupported {
 			return nil, err // TODO go1.13 use wrapping
@@ -150,65 +151,17 @@ func MoveReplicationCursor(ctx context.Context, fs string, target ReplicationCur
 		return nil, errors.Wrapf(err, "cannot create bookmark")
 	}
 
-	destroyedCursors, err = DestroyObsoleteReplicationCursors(ctx, fs, target, jobID)
-	if err != nil {
-		return nil, errors.Wrap(err, "destroy obsolete replication cursors")
-	}
-
-	return destroyedCursors, nil
+	return &bookmarkBasedAbstraction{
+		Type:              AbstractionReplicationCursorBookmarkV2,
+		FS:                fs,
+		FilesystemVersion: cursorBookmark,
+		JobID:             jobID,
+	}, nil
 }
 
-type ReplicationCursor interface {
-	GetCreateTXG() uint64
-}
-
-func DestroyObsoleteReplicationCursors(ctx context.Context, fs string, current ReplicationCursor, jobID JobID) (_ []Abstraction, err error) {
-
-	q := ListZFSHoldsAndBookmarksQuery{
-		FS: ListZFSHoldsAndBookmarksQueryFilesystemFilter{
-			FS: &fs,
-		},
-		What: AbstractionTypeSet{
-			AbstractionReplicationCursorBookmarkV2: true,
-		},
-		JobID: &jobID,
-		CreateTXG: CreateTXGRange{
-			Since: nil,
-			Until: &CreateTXGRangeBound{
-				CreateTXG: current.GetCreateTXG(),
-				Inclusive: &zfs.NilBool{B: false},
-			},
-		},
-		Concurrency: 1,
-	}
-	abs, absErr, err := ListAbstractions(ctx, q)
-	if err != nil {
-		return nil, errors.Wrap(err, "list abstractions")
-	}
-	if len(absErr) > 0 {
-		return nil, errors.Wrap(ListAbstractionsErrors(absErr), "list abstractions")
-	}
-
-	var destroyed []Abstraction
-	var errs []error
-	for res := range BatchDestroy(ctx, abs) {
-		log := getLogger(ctx).
-			WithField("replication_cursor_bookmark", res.Abstraction)
-		if res.DestroyErr != nil {
-			errs = append(errs, res.DestroyErr)
-			log.WithError(err).
-				Error("cannot destroy obsolete replication cursor bookmark")
-		} else {
-			destroyed = append(destroyed, res.Abstraction)
-			log.Info("destroyed obsolete replication cursor bookmark")
-		}
-	}
-	if len(errs) == 0 {
-		return destroyed, nil
-	} else {
-		return destroyed, errorarray.Wrap(errs, "destroy obsolete replication cursor")
-	}
-}
+const (
+	ReplicationCursorBookmarkNamePrefix = "zrepl_last_received_J_"
+)
 
 var lastReceivedHoldTagRE = regexp.MustCompile("^zrepl_last_received_J_(.+)$")
 
@@ -230,22 +183,22 @@ func LastReceivedHoldTag(jobID JobID) (string, error) {
 }
 
 func lastReceivedHoldImpl(jobid string) (string, error) {
-	tag := fmt.Sprintf("zrepl_last_received_J_%s", jobid)
+	tag := fmt.Sprintf("%s%s", ReplicationCursorBookmarkNamePrefix, jobid)
 	if err := zfs.ValidHoldTag(tag); err != nil {
 		return "", err
 	}
 	return tag, nil
 }
 
-func MoveLastReceivedHold(ctx context.Context, fs string, to zfs.FilesystemVersion, jobID JobID) error {
+func CreateLastReceivedHold(ctx context.Context, fs string, to zfs.FilesystemVersion, jobID JobID) (Abstraction, error) {
 
 	if !to.IsSnapshot() {
-		return errors.Errorf("last-received-hold: target must be a snapshot: %s", to.FullPath(fs))
+		return nil, errors.Errorf("last-received-hold: target must be a snapshot: %s", to.FullPath(fs))
 	}
 
 	tag, err := LastReceivedHoldTag(jobID)
 	if err != nil {
-		return errors.Wrap(err, "last-received-hold: hold tag")
+		return nil, errors.Wrap(err, "last-received-hold: hold tag")
 	}
 
 	// we never want to be without a hold
@@ -253,7 +206,23 @@ func MoveLastReceivedHold(ctx context.Context, fs string, to zfs.FilesystemVersi
 
 	err = zfs.ZFSHold(ctx, fs, to, tag)
 	if err != nil {
-		return errors.Wrap(err, "last-received-hold: hold newly received")
+		return nil, errors.Wrap(err, "last-received-hold: hold newly received")
+	}
+
+	return &holdBasedAbstraction{
+		Type:              AbstractionLastReceivedHold,
+		FS:                fs,
+		FilesystemVersion: to,
+		JobID:             jobID,
+		Tag:               tag,
+	}, nil
+}
+
+func MoveLastReceivedHold(ctx context.Context, fs string, to zfs.FilesystemVersion, jobID JobID) error {
+
+	_, err := CreateLastReceivedHold(ctx, fs, to, jobID)
+	if err != nil {
+		return err
 	}
 
 	q := ListZFSHoldsAndBookmarksQuery{

endpoint/endpoint_zfs_abstraction_pdu.go

@@ -0,0 +1,28 @@
+package endpoint
+
+import "github.com/zrepl/zrepl/replication/logic/pdu"
+
+func SendAbstractionToPDU(a Abstraction) *pdu.SendAbstraction {
+	var ty pdu.SendAbstraction_SendAbstractionType
+	switch a.GetType() {
+	case AbstractionLastReceivedHold:
+		panic(a)
+	case AbstractionReplicationCursorBookmarkV1:
+		panic(a)
+	case AbstractionReplicationCursorBookmarkV2:
+		ty = pdu.SendAbstraction_ReplicationCursorV2
+	case AbstractionStepHold:
+		ty = pdu.SendAbstraction_StepHold
+	case AbstractionStepBookmark:
+		ty = pdu.SendAbstraction_StepBookmark
+	default:
+		panic(a)
+	}
+
+	version := a.GetFilesystemVersion()
+	return &pdu.SendAbstraction{
+		Type:    ty,
+		JobID:   (*a.GetJobID()).String(),
+		Version: pdu.FilesystemVersionFromZFS(&version),
+	}
+}

endpoint/endpoint_zfs_abstraction_step.go

@@ -7,7 +7,6 @@ import (
 
 	"github.com/pkg/errors"
 
-	"github.com/zrepl/zrepl/util/errorarray"
 	"github.com/zrepl/zrepl/zfs"
 )
 
@@ -93,7 +92,7 @@ func HoldStep(ctx context.Context, fs string, v zfs.FilesystemVersion, jobID Job
 		return nil, errors.Wrap(err, "create step bookmark: determine bookmark name")
 	}
 	// idempotently create bookmark
-	err = zfs.ZFSBookmark(ctx, fs, v.ToSendArgVersion(), bmname)
+	stepBookmark, err := zfs.ZFSBookmark(ctx, fs, v, bmname)
 	if err != nil {
 		if err == zfs.ErrBookmarkCloningNotSupported {
 			// TODO we could actually try to find a local snapshot that has the requested GUID
@@ -108,137 +107,11 @@ func HoldStep(ctx context.Context, fs string, v zfs.FilesystemVersion, jobID Job
 	return &bookmarkBasedAbstraction{
 		Type:              AbstractionStepBookmark,
 		FS:                fs,
-		FilesystemVersion: v,
+		FilesystemVersion: stepBookmark,
 		JobID:             jobID,
 	}, nil
 }
 
-// idempotently release the step-hold on v if v is a snapshot
-// or idempotently destroy the step-bookmark  of v if v is a bookmark
-//
-// note that this operation leaves v itself untouched, unless v is the step-bookmark itself, in which case v is destroyed
-//
-// returns an instance of *zfs.DatasetDoesNotExist if `v` does not exist
-func ReleaseStep(ctx context.Context, fs string, v zfs.FilesystemVersion, jobID JobID) error {
-
-	if v.IsSnapshot() {
-		tag, err := StepHoldTag(jobID)
-		if err != nil {
-			return errors.Wrap(err, "step release tag")
-		}
-
-		if err := zfs.ZFSRelease(ctx, tag, v.FullPath(fs)); err != nil {
-			return errors.Wrap(err, "step release: zfs")
-		}
-
-		return nil
-	}
-	if !v.IsBookmark() {
-		panic(fmt.Sprintf("impl error: expecting version to be a bookmark, got %#v", v))
-	}
-
-	bmname, err := StepBookmarkName(fs, v.Guid, jobID)
-	if err != nil {
-		return errors.Wrap(err, "step release: determine bookmark name")
-	}
-	// idempotently destroy bookmark
-
-	if err := zfs.ZFSDestroyIdempotent(ctx, bmname); err != nil {
-		return errors.Wrap(err, "step release: bookmark destroy: zfs")
-	}
-
-	return nil
-}
-
-// release {step holds, step bookmarks} earlier and including `mostRecent`
-func ReleaseStepCummulativeInclusive(ctx context.Context, fs string, since *CreateTXGRangeBound, mostRecent zfs.FilesystemVersion, jobID JobID) error {
-	q := ListZFSHoldsAndBookmarksQuery{
-		What: AbstractionTypeSet{
-			AbstractionStepHold:     true,
-			AbstractionStepBookmark: true,
-		},
-		FS: ListZFSHoldsAndBookmarksQueryFilesystemFilter{
-			FS: &fs,
-		},
-		JobID: &jobID,
-		CreateTXG: CreateTXGRange{
-			Since: since,
-			Until: &CreateTXGRangeBound{
-				CreateTXG: mostRecent.CreateTXG,
-				Inclusive: &zfs.NilBool{B: true},
-			},
-		},
-		Concurrency: 1,
-	}
-	abs, absErrs, err := ListAbstractions(ctx, q)
-	if err != nil {
-		return errors.Wrap(err, "step release cummulative: list")
-	}
-	if len(absErrs) > 0 {
-		return errors.Wrap(ListAbstractionsErrors(absErrs), "step release cummulative: list")
-	}
-
-	getLogger(ctx).WithField("step_holds_and_bookmarks", fmt.Sprintf("%s", abs)).Debug("releasing step holds and bookmarks")
-
-	var errs []error
-	for res := range BatchDestroy(ctx, abs) {
-		log := getLogger(ctx).
-			WithField("step_hold_or_bookmark", res.Abstraction)
-		if res.DestroyErr != nil {
-			errs = append(errs, res.DestroyErr)
-			log.WithError(err).
-				Error("cannot release step hold or bookmark")
-		} else {
-			log.Info("released step hold or bookmark")
-		}
-	}
-	if len(errs) == 0 {
-		return nil
-	} else {
-		return errorarray.Wrap(errs, "step release cummulative: release")
-	}
-}
-
-func TryReleaseStepStaleFS(ctx context.Context, fs string, jobID JobID) {
-
-	q := ListZFSHoldsAndBookmarksQuery{
-		FS: ListZFSHoldsAndBookmarksQueryFilesystemFilter{
-			FS: &fs,
-		},
-		JobID: &jobID,
-		What: AbstractionTypeSet{
-			AbstractionStepHold:                    true,
-			AbstractionStepBookmark:                true,
-			AbstractionReplicationCursorBookmarkV2: true,
-		},
-		Concurrency: 1,
-	}
-	staleness, err := ListStale(ctx, q)
-	if _, ok := err.(*ListStaleQueryError); ok {
-		panic(err)
-	} else if err != nil {
-		getLogger(ctx).WithError(err).Error("cannot list stale step holds and bookmarks")
-		return
-	}
-	for _, s := range staleness.Stale {
-		getLogger(ctx).WithField("stale_step_hold_or_bookmark", s).Info("batch-destroying stale step hold or bookmark")
-	}
-	for res := range BatchDestroy(ctx, staleness.Stale) {
-		if res.DestroyErr != nil {
-			getLogger(ctx).
-				WithField("stale_step_hold_or_bookmark", res.Abstraction).
-				WithError(res.DestroyErr).
-				Error("cannot destroy stale step-hold or bookmark")
-		} else {
-			getLogger(ctx).
-				WithField("stale_step_hold_or_bookmark", res.Abstraction).
-				WithError(res.DestroyErr).
-				Info("destroyed stale step-hold or bookmark")
-		}
-	}
-
-}
-
 var _ BookmarkExtractor = StepBookmarkExtractor
 
 func StepBookmarkExtractor(fs *zfs.DatasetPath, v zfs.FilesystemVersion) (_ Abstraction) {

go.mod

@@ -5,6 +5,7 @@ go 1.12
 require (
 	github.com/fatih/color v1.7.0
 	github.com/gdamore/tcell v1.2.0
+	github.com/gitchander/permutation v0.0.0-20181107151852-9e56b92e9909
 	github.com/go-logfmt/logfmt v0.4.0
 	github.com/go-sql-driver/mysql v1.4.1-0.20190907122137-b2c03bcae3d4
 	github.com/golang/protobuf v1.3.2
@@ -23,16 +24,19 @@ require (
 	github.com/pkg/profile v1.2.1
 	github.com/problame/go-netssh v0.0.0-20191209123953-18d8aa6923c7
 	github.com/prometheus/client_golang v1.2.1
+	github.com/prometheus/common v0.7.0
 	github.com/sergi/go-diff v1.0.1-0.20180205163309-da645544ed44 // go1.12 thinks it needs this
 	github.com/spf13/cobra v0.0.2
 	github.com/spf13/pflag v1.0.5
 	github.com/stretchr/testify v1.4.0
+	github.com/willf/bitset v1.1.10
 	github.com/yudai/gojsondiff v0.0.0-20170107030110-7b1b7adf999d
 	github.com/yudai/golcs v0.0.0-20170316035057-ecda9a501e82 // go1.12 thinks it needs this
 	github.com/zrepl/yaml-config v0.0.0-20191220194647-cbb6b0cf4bdd
 	golang.org/x/net v0.0.0-20190613194153-d28f0bde5980
 	golang.org/x/sync v0.0.0-20190423024810-112230192c58
 	golang.org/x/sys v0.0.0-20191026070338-33540a1f6037
+	golang.org/x/tools v0.0.0-20190206041539-40960b6deb8e
 	gonum.org/v1/gonum v0.7.0 // indirect
 	google.golang.org/grpc v1.17.0
 )

go.sum

@@ -7,8 +7,10 @@ github.com/OpenPeeDeeP/depguard v0.0.0-20181229194401-1f388ab2d810/go.mod h1:7/4
 github.com/StackExchange/wmi v0.0.0-20180116203802-5d049714c4a6/go.mod h1:3eOhrUMpNV+6aFIbp5/iudMxNCF27Vw2OZgy4xEx0Fg=
 github.com/ajstarks/svgo v0.0.0-20180226025133-644b8db467af/go.mod h1:K08gAheRH3/J6wwsYMMT4xOr94bZjxIelGM0+d/wbFw=
 github.com/alecthomas/template v0.0.0-20160405071501-a0175ee3bccc/go.mod h1:LOuyumcjzFXgccqObfd/Ljyb9UuFJ6TxHnclSeseNhc=
+github.com/alecthomas/template v0.0.0-20190718012654-fb15b899a751 h1:JYp7IbQjafoB+tBA3gMyHYHrpOtNuDiK/uB5uXxq5wM=
 github.com/alecthomas/template v0.0.0-20190718012654-fb15b899a751/go.mod h1:LOuyumcjzFXgccqObfd/Ljyb9UuFJ6TxHnclSeseNhc=
 github.com/alecthomas/units v0.0.0-20151022065526-2efee857e7cf/go.mod h1:ybxpYRFXyAe+OPACYpWeL0wqObRcbAqCMya13uyzqw0=
+github.com/alecthomas/units v0.0.0-20190717042225-c3de453c63f4 h1:Hs82Z41s6SdL1CELW+XaDYmOH4hkBN4/N9og/AsOv7E=
 github.com/alecthomas/units v0.0.0-20190717042225-c3de453c63f4/go.mod h1:ybxpYRFXyAe+OPACYpWeL0wqObRcbAqCMya13uyzqw0=
 github.com/alvaroloes/enumer v1.1.1/go.mod h1:FxrjvuXoDAx9isTJrv4c+T410zFi0DtXIT0m65DJ+Wo=
 github.com/armon/consul-api v0.0.0-20180202201655-eb2c6b5be1b6/go.mod h1:grANhF5doyWs3UAsr3K4I6qtAmlQcZDesFNEHPZAzj8=
@@ -38,6 +40,8 @@ github.com/gdamore/encoding v1.0.0 h1:+7OoQ1Bc6eTm5niUzBa0Ctsh6JbMW6Ra+YNuAtDBdk
 github.com/gdamore/encoding v1.0.0/go.mod h1:alR0ol34c49FCSBLjhosxzcPHQbf2trDkoo5dl+VrEg=
 github.com/gdamore/tcell v1.2.0 h1:ikixzsxc8K8o3V2/CEmyoEW8mJZaNYQQ3NP3VIQdUe4=
 github.com/gdamore/tcell v1.2.0/go.mod h1:Hjvr+Ofd+gLglo7RYKxxnzCBmev3BzsS67MebKS4zMM=
+github.com/gitchander/permutation v0.0.0-20181107151852-9e56b92e9909 h1:9NC8seTx6/zRmMTAdsHj/uOMi0EGHGQtjyLafBjk77Q=
+github.com/gitchander/permutation v0.0.0-20181107151852-9e56b92e9909/go.mod h1:lP+DW8LR6Rw3ru9Vo2/y/3iiLaLWmofYql/va+7zJOk=
 github.com/go-critic/go-critic v0.3.4/go.mod h1:AHR42Lk/E/aOznsrYdMYeIQS5RH10HZHSqP+rD6AJrc=
 github.com/go-critic/go-critic v0.3.5-0.20190526074819-1df300866540/go.mod h1:+sE8vrLDS2M0pZkBk0wy6+nLdKexVDrl/jBqQOTDThA=
 github.com/go-kit/kit v0.8.0/go.mod h1:xBxKIO96dXMWWy0MnWVtmwkA9/13aqxPnvrjFYMA2as=
@@ -242,6 +246,7 @@ github.com/shurcooL/go-goon v0.0.0-20170922171312-37c2f522c041/go.mod h1:N5mDOms
 github.com/sirupsen/logrus v1.0.5/go.mod h1:pMByvHTf9Beacp5x1UXfOR9xyW/9antXMhjMPG0dEzc=
 github.com/sirupsen/logrus v1.2.0/go.mod h1:LxeOpSwHxABJmUn/MG1IvRgCAasNZTLOkJPxbbu5VWo=
 github.com/sirupsen/logrus v1.4.0/go.mod h1:LxeOpSwHxABJmUn/MG1IvRgCAasNZTLOkJPxbbu5VWo=
+github.com/sirupsen/logrus v1.4.2 h1:SPIRibHv4MatM3XXNO2BJeFLZwZ2LvZgfQ5+UNI2im4=
 github.com/sirupsen/logrus v1.4.2/go.mod h1:tLMulIdttU9McNUspp0xgXVQah82FyeX6MwdIuYE2rE=
 github.com/sourcegraph/go-diff v0.5.1/go.mod h1:j2dHj3m8aZgQO8lMTcTnBcXkRRRqi34cd2MNlA9u1mE=
 github.com/spf13/afero v1.1.0/go.mod h1:j4pytiNVoe2o6bmDsKpLACNPDBIoEAkihy7loJ1B0CQ=
@@ -263,6 +268,7 @@ github.com/spf13/viper v1.0.2/go.mod h1:A8kyI5cUJhb8N+3pkfONlcEcZbueH6nhAm0Fq7Sr
 github.com/spf13/viper v1.3.2/go.mod h1:ZiWeW+zYFKm7srdB9IoDzzZXaJaI5eL9QjNiN/DMA2s=
 github.com/stretchr/objx v0.1.0/go.mod h1:HFkY916IF+rwdDfMAkV7OtwuqBVzrE8GR6GFx+wExME=
 github.com/stretchr/objx v0.1.1/go.mod h1:HFkY916IF+rwdDfMAkV7OtwuqBVzrE8GR6GFx+wExME=
+github.com/stretchr/testify v1.2.1/go.mod h1:a8OnRcib4nhh0OaRAV+Yts87kKdq0PP7pXfy6kDkUVs=
 github.com/stretchr/testify v1.2.2 h1:bSDNvY7ZPG5RlJ8otE/7V6gMiyenm9RtJ7IUVIAoJ1w=
 github.com/stretchr/testify v1.2.2/go.mod h1:a8OnRcib4nhh0OaRAV+Yts87kKdq0PP7pXfy6kDkUVs=
 github.com/stretchr/testify v1.3.0/go.mod h1:M5WIy9Dh21IEIfnGCwXGc5bZfKNJtfHm1UVUgZn+9EI=
@@ -276,6 +282,8 @@ github.com/valyala/bytebufferpool v1.0.0/go.mod h1:6bBcMArwyJ5K/AmCkWv1jt77kVWyC
 github.com/valyala/fasthttp v1.2.0/go.mod h1:4vX61m6KN+xDduDNwXrhIAVZaZaZiQ1luJk8LWSxF3s=
 github.com/valyala/quicktemplate v1.1.1/go.mod h1:EH+4AkTd43SvgIbQHYu59/cJyxDoOVRUAfrukLPuGJ4=
 github.com/valyala/tcplisten v0.0.0-20161114210144-ceec8f93295a/go.mod h1:v3UYOV9WzVtRmSR+PDvWpU/qWl4Wa5LApYYX4ZtKbio=
+github.com/willf/bitset v1.1.10 h1:NotGKqX0KwQ72NUzqrjZq5ipPNDQex9lo3WpaS8L2sc=
+github.com/willf/bitset v1.1.10/go.mod h1:RjeCKbqT1RxIR/KWY6phxZiaY1IyutSBfGjNPySAYV4=
 github.com/xordataexchange/crypt v0.0.3-0.20170626215501-b2862e3d0a77/go.mod h1:aYKd//L2LvnjZzWKhF00oedf4jCCReLcmhLdhm1A27Q=
 github.com/yudai/gojsondiff v0.0.0-20170107030110-7b1b7adf999d h1:yJIizrfO599ot2kQ6Af1enICnwBD3XoxgX3MrMwot2M=
 github.com/yudai/gojsondiff v0.0.0-20170107030110-7b1b7adf999d/go.mod h1:AY32+k2cwILAkW1fbgxQ5mUmMiZFgLIV+FBNExI05xg=
@@ -343,6 +351,7 @@ golang.org/x/tools v0.0.0-20181117154741-2ddaf7f79a09/go.mod h1:n7NCudcB/nEzxVGm
 golang.org/x/tools v0.0.0-20181205014116-22934f0fdb62/go.mod h1:n7NCudcB/nEzxVGmLbDWY5pfWTLqBcC2KZ6jyYvM4mQ=
 golang.org/x/tools v0.0.0-20190110163146-51295c7ec13a/go.mod h1:n7NCudcB/nEzxVGmLbDWY5pfWTLqBcC2KZ6jyYvM4mQ=
 golang.org/x/tools v0.0.0-20190121143147-24cd39ecf745/go.mod h1:n7NCudcB/nEzxVGmLbDWY5pfWTLqBcC2KZ6jyYvM4mQ=
+golang.org/x/tools v0.0.0-20190206041539-40960b6deb8e h1:Io7mpb+aUAGF0MKxbyQ7HQl1VgB+cL6ZJZUFaFNqVV4=
 golang.org/x/tools v0.0.0-20190206041539-40960b6deb8e/go.mod h1:n7NCudcB/nEzxVGmLbDWY5pfWTLqBcC2KZ6jyYvM4mQ=
 golang.org/x/tools v0.0.0-20190213192042-740235f6c0d8/go.mod h1:n7NCudcB/nEzxVGmLbDWY5pfWTLqBcC2KZ6jyYvM4mQ=
 golang.org/x/tools v0.0.0-20190311212946-11955173bddd/go.mod h1:LCzVGOaR6xXOjkQ3onu1FJEFr0SW1gC7cKk1uF8kGRs=
@@ -365,6 +374,7 @@ google.golang.org/grpc v1.16.0/go.mod h1:0JHn/cJsOMiMfNA9+DeHDlAU7KAAB5GDlYFpa9M
 google.golang.org/grpc v1.17.0 h1:TRJYBgMclJvGYn2rIMjj+h9KtMt5r1Ij7ODVRIZkwhk=
 google.golang.org/grpc v1.17.0/go.mod h1:6QZJwpn2B+Zp71q/5VxRsJ6NXXVCE5NRUHRo+f3cWCs=
 gopkg.in/airbrake/gobrake.v2 v2.0.9/go.mod h1:/h5ZAUhDkGaJfjzjKLSjv6zCL6O0LLBxU4K+aSYdM/U=
+gopkg.in/alecthomas/kingpin.v2 v2.2.6 h1:jMFz6MfLP0/4fUyZle81rXUoxOBFi19VUFKVDOQfozc=
 gopkg.in/alecthomas/kingpin.v2 v2.2.6/go.mod h1:FMv+mEhP44yOT+4EoQTLFTRgOQ1FBLkstjWtayDeSgw=
 gopkg.in/check.v1 v0.0.0-20161208181325-20d25e280405 h1:yhCVgyC4o1eVCa2tZl7eS0r+SDo693bJlVdllGtEeKM=
 gopkg.in/check.v1 v0.0.0-20161208181325-20d25e280405/go.mod h1:Co6ibVJAznAaIkqp8huTwlJQCZ016jof/cbN4VW5Yz0=

lazy.sh

@@ -36,9 +36,10 @@ godep() {
     go build -v -mod=readonly -o "$GOPATH/bin/enumer"        github.com/alvaroloes/enumer
     go build -v -mod=readonly -o "$GOPATH/bin/goimports"     golang.org/x/tools/cmd/goimports
     go build -v -mod=readonly -o "$GOPATH/bin/golangci-lint" github.com/golangci/golangci-lint/cmd/golangci-lint
+    go build -v -mod=readonly -o "$GOPATH/bin/gocovmerge"    github.com/wadey/gocovmerge
     set +x
     popd
-    if ! type stringer || ! type protoc-gen-go || ! type enumer || ! type goimports || ! type golangci-lint; then
+    if ! type stringer || ! type protoc-gen-go || ! type enumer || ! type goimports || ! type golangci-lint || ! type gocovmerge; then
         echo "Installed dependencies but can't find them in \$PATH, adjust it to contain \$GOPATH/bin" 1>&2
         exit 1
     fi

main.go

@@ -17,7 +17,7 @@ func init() {
 	cli.AddSubcommand(client.PprofCmd)
 	cli.AddSubcommand(client.TestCmd)
 	cli.AddSubcommand(client.MigrateCmd)
-	cli.AddSubcommand(client.HoldsCmd)
+	cli.AddSubcommand(client.ZFSAbstractionsCmd)
 }
 
 func main() {

platformtest/harness/harness.go

@@ -11,6 +11,7 @@ import (
 
 	"github.com/fatih/color"
 	"github.com/pkg/errors"
+	"github.com/zrepl/zrepl/daemon/logging/trace"
 
 	"github.com/zrepl/zrepl/config"
 	"github.com/zrepl/zrepl/daemon/logging"
@@ -23,33 +24,36 @@ var bold = color.New(color.Bold)
 var boldRed = color.New(color.Bold, color.FgHiRed)
 var boldGreen = color.New(color.Bold, color.FgHiGreen)
 
-var args struct {
-	createArgs            platformtest.ZpoolCreateArgs
-	stopAndKeepPoolOnFail bool
-
-	run   string
-	runRE *regexp.Regexp
-}
+const DefaultPoolImageSize = 200 * (1 << 20)
 
 func main() {
-	if err := doMain(); err != nil {
+
+	var args HarnessArgs
+
+	flag.StringVar(&args.CreateArgs.PoolName, "poolname", "", "")
+	flag.StringVar(&args.CreateArgs.ImagePath, "imagepath", "", "")
+	flag.Int64Var(&args.CreateArgs.ImageSize, "imagesize", DefaultPoolImageSize, "")
+	flag.StringVar(&args.CreateArgs.Mountpoint, "mountpoint", "", "")
+	flag.BoolVar(&args.StopAndKeepPoolOnFail, "failure.stop-and-keep-pool", false, "if a test case fails, stop test execution and keep pool as it was when the test failed")
+	flag.StringVar(&args.Run, "run", "", "")
+	flag.Parse()
+
+	if err := HarnessRun(args); err != nil {
 		os.Exit(1)
 	}
 }
 
 var exitWithErr = fmt.Errorf("exit with error")
 
-func doMain() error {
+type HarnessArgs struct {
+	CreateArgs            platformtest.ZpoolCreateArgs
+	StopAndKeepPoolOnFail bool
+	Run                   string
+}
 
-	flag.StringVar(&args.createArgs.PoolName, "poolname", "", "")
-	flag.StringVar(&args.createArgs.ImagePath, "imagepath", "", "")
-	flag.Int64Var(&args.createArgs.ImageSize, "imagesize", 200*(1<<20), "")
-	flag.StringVar(&args.createArgs.Mountpoint, "mountpoint", "", "")
-	flag.BoolVar(&args.stopAndKeepPoolOnFail, "failure.stop-and-keep-pool", false, "if a test case fails, stop test execution and keep pool as it was when the test failed")
-	flag.StringVar(&args.run, "run", "", "")
-	flag.Parse()
+func HarnessRun(args HarnessArgs) error {
 
-	args.runRE = regexp.MustCompile(args.run)
+	runRE := regexp.MustCompile(args.Run)
 
 	outlets := logger.NewOutlets()
 	outlet, level, err := logging.ParseOutlet(config.LoggingOutletEnum{Ret: &config.StdoutLoggingOutlet{
@@ -64,11 +68,13 @@ func doMain() error {
 	outlets.Add(outlet, level)
 	logger := logger.NewLogger(outlets, 1*time.Second)
 
-	if err := args.createArgs.Validate(); err != nil {
+	if err := args.CreateArgs.Validate(); err != nil {
 		logger.Error(err.Error())
 		panic(err)
 	}
-	ctx := platformtest.WithLogger(context.Background(), logger)
+	ctx := context.Background()
+	defer trace.WithTaskFromStackUpdateCtx(&ctx)()
+	ctx = logging.WithLoggers(ctx, logging.SubsystemLoggersWithUniversalLogger(logger))
 	ex := platformtest.NewEx(logger)
 
 	type invocation struct {
@@ -78,7 +84,7 @@ func doMain() error {
 
 	invocations := make([]*invocation, 0, len(tests.Cases))
 	for _, c := range tests.Cases {
-		if args.runRE.MatchString(c.String()) {
+		if runRE.MatchString(c.String()) {
 			invocations = append(invocations, &invocation{runFunc: c})
 		}
 	}
@@ -87,7 +93,7 @@ func doMain() error {
 
 		bold.Printf("BEGIN TEST CASE %s\n", inv.runFunc.String())
 
-		pool, err := platformtest.CreateOrReplaceZpool(ctx, ex, args.createArgs)
+		pool, err := platformtest.CreateOrReplaceZpool(ctx, ex, args.CreateArgs)
 		if err != nil {
 			panic(errors.Wrap(err, "create test pool"))
 		}
@@ -103,7 +109,7 @@ func doMain() error {
 			fmt.Printf("%+v\n", res.failedStack) // print with stack trace
 		}
 
-		if res.failed && args.stopAndKeepPoolOnFail {
+		if res.failed && args.StopAndKeepPoolOnFail {
 			boldRed.Printf("STOPPING TEST RUN AT FAILING TEST PER USER REQUEST\n")
 			return exitWithErr
 		}

platformtest/harness/harness_test.go

@@ -1,6 +1,7 @@
 package main
 
 import (
+	"fmt"
 	"os"
 	"strings"
 	"testing"
@@ -21,22 +22,29 @@ import (
 //  https://github.com/wadey/gocovmerge
 
 func TestMain(t *testing.T) {
+	fmt.Println("incoming args: ", os.Args)
+
 	var (
-		args []string
-		run  bool
+		args             []string
+		run              bool
+		startCaptureArgs bool
 	)
 
-	for _, arg := range os.Args {
+	for i, arg := range os.Args {
 		switch {
 		case arg == "__DEVEL--i-heard-you-like-tests":
 			run = true
+			startCaptureArgs = true
 		case strings.HasPrefix(arg, "-test"):
 		case strings.HasPrefix(arg, "__DEVEL"):
-		default:
+		case i == 0:
+			args = append(args, arg)
+		case startCaptureArgs:
 			args = append(args, arg)
 		}
 	}
 	os.Args = args
+	fmt.Println("using args: ", os.Args)
 
 	if run {
 		main()

platformtest/platformtest.go

@@ -27,6 +27,7 @@ func (c *Context) Logf(format string, args ...interface{}) {
 
 func (c *Context) Errorf(format string, args ...interface{}) {
 	GetLog(c).Printf(format, args...)
+	c.FailNow()
 }
 
 func (c *Context) FailNow() {

platformtest/platformtest_logging.go

@@ -3,22 +3,12 @@ package platformtest
 import (
 	"context"
 
+	"github.com/zrepl/zrepl/daemon/logging"
 	"github.com/zrepl/zrepl/logger"
 )
 
 type Logger = logger.Logger
 
-type contextKey int
-
-const (
-	contextKeyLogger contextKey = iota
-)
-
-func WithLogger(ctx context.Context, logger Logger) context.Context {
-	ctx = context.WithValue(ctx, contextKeyLogger, logger)
-	return ctx
-}
-
 func GetLog(ctx context.Context) Logger {
-	return ctx.Value(contextKeyLogger).(Logger)
+	return logging.GetLogger(ctx, logging.SubsysPlatformtest)
 }

platformtest/platformtest_zpool.go

@@ -58,6 +58,14 @@ func CreateOrReplaceZpool(ctx context.Context, e Execer, args ZpoolCreateArgs) (
 		}
 	}
 
+	// clear the mountpoint dir
+	if err := os.RemoveAll(args.Mountpoint); err != nil {
+		return nil, errors.Wrapf(err, "remove mountpoint dir %q", args.Mountpoint)
+	}
+	if err := os.Mkdir(args.Mountpoint, 0700); err != nil {
+		return nil, errors.Wrapf(err, "create mountpoint dir %q", args.Mountpoint)
+	}
+
 	// idempotently (re)create the pool image
 	image, err := os.OpenFile(args.ImagePath, os.O_CREATE|os.O_RDWR, 0600)
 	if err != nil {
@@ -93,5 +101,9 @@ func (p *Zpool) Destroy(ctx context.Context, e Execer) error {
 		return errors.Wrapf(err, "remove pool image")
 	}
 
+	if err := os.RemoveAll(p.args.Mountpoint); err != nil {
+		return errors.Wrapf(err, "remove mountpoint dir %q", p.args.Mountpoint)
+	}
+
 	return nil
 }

platformtest/tests/batchDestroy.go

@@ -2,8 +2,9 @@ package tests
 
 import (
 	"fmt"
-	"strings"
 
+	"github.com/kr/pretty"
+	"github.com/stretchr/testify/require"
 	"github.com/zrepl/zrepl/platformtest"
 	"github.com/zrepl/zrepl/zfs"
 )
@@ -17,7 +18,9 @@ func BatchDestroy(ctx *platformtest.Context) {
 		+  "foo bar@1"
 		+  "foo bar@2"
 		+  "foo bar@3"
+		+  "foo bar@4"
 		R  zfs hold zrepl_platformtest "${ROOTDS}/foo bar@2"
+		R  zfs hold zrepl_platformtest "${ROOTDS}/foo bar@4"
 	`)
 
 	reqs := []*zfs.DestroySnapOp{
@@ -31,24 +34,40 @@ func BatchDestroy(ctx *platformtest.Context) {
 			Filesystem: fmt.Sprintf("%s/foo bar", ctx.RootDataset),
 			Name:       "2",
 		},
+		&zfs.DestroySnapOp{
+			ErrOut:     new(error),
+			Filesystem: fmt.Sprintf("%s/foo bar", ctx.RootDataset),
+			Name:       "non existent",
+		},
+		&zfs.DestroySnapOp{
+			ErrOut:     new(error),
+			Filesystem: fmt.Sprintf("%s/foo bar", ctx.RootDataset),
+			Name:       "4",
+		},
 	}
-	zfs.ZFSDestroyFilesystemVersions(reqs)
+	zfs.ZFSDestroyFilesystemVersions(ctx, reqs)
+
+	pretty.Println(reqs)
+
 	if *reqs[0].ErrOut != nil {
 		panic("expecting no error")
 	}
-	err := (*reqs[1].ErrOut).Error()
-	if !strings.Contains(err, fmt.Sprintf("%s/foo bar@2", ctx.RootDataset)) {
-		panic(fmt.Sprintf("expecting error about being unable to destroy @2: %T\n%s", err, err))
-	}
+
+	eBusy, ok := (*reqs[1].ErrOut).(*zfs.ErrDestroySnapshotDatasetIsBusy)
+	require.True(ctx, ok)
+	require.Equal(ctx, reqs[1].Name, eBusy.Name)
+
+	require.Nil(ctx, *reqs[2].ErrOut, "destroying non-existent snap is not an error (idempotence)")
+
+	eBusy, ok = (*reqs[3].ErrOut).(*zfs.ErrDestroySnapshotDatasetIsBusy)
+	require.True(ctx, ok)
+	require.Equal(ctx, reqs[3].Name, eBusy.Name)
 
 	platformtest.Run(ctx, platformtest.PanicErr, ctx.RootDataset, `
 	!N "foo bar@3"
 	!E "foo bar@1"
 	!E "foo bar@2"
-	R zfs release zrepl_platformtest "${ROOTDS}/foo bar@2"
-	-  "foo bar@2"
-	-  "foo bar@1"
-	-  "foo bar"
+	!E "foo bar@4"
 	`)
 
 }

platformtest/tests/gen/gen.go

@@ -0,0 +1,140 @@
+package main
+
+import (
+	"bytes"
+	"go/ast"
+	"go/format"
+	"go/parser"
+	"go/token"
+	"io/ioutil"
+	"os"
+	"sort"
+	"strings"
+	"text/template"
+
+	"golang.org/x/tools/go/packages"
+)
+
+func check(err error) {
+	if err != nil {
+		panic(err)
+	}
+}
+
+type platformtestFuncDeclFinder struct {
+	pkg       *packages.Package
+	testFuncs []*ast.FuncDecl
+}
+
+func isPlatformtestFunc(n *ast.FuncDecl) bool {
+	if !n.Name.IsExported() {
+		return false
+	}
+	if n.Recv != nil {
+		return false
+	}
+	if n.Type.Results.NumFields() != 0 {
+		return false
+	}
+	if n.Type.Params.NumFields() != 1 {
+		return false
+	}
+	se, ok := n.Type.Params.List[0].Type.(*ast.StarExpr)
+	if !ok {
+		return false
+	}
+	sel, ok := se.X.(*ast.SelectorExpr)
+	if !ok {
+		return false
+	}
+	x, ok := sel.X.(*ast.Ident)
+	if !ok {
+		return false
+	}
+	if x.Name != "platformtest" || sel.Sel.Name != "Context" {
+		return false
+	}
+	return true
+}
+
+func (e *platformtestFuncDeclFinder) Visit(n2 ast.Node) ast.Visitor {
+	switch n := n2.(type) {
+	case *ast.File:
+		return e
+	case *ast.FuncDecl:
+		if isPlatformtestFunc(n) {
+			e.testFuncs = append(e.testFuncs, n)
+		}
+		return nil
+	default:
+		return nil
+	}
+}
+
+func main() {
+
+	// TODO safeguards that prevent us from deleting non-generated generated_cases.go
+	os.Remove("generated_cases.go")
+	// (no error handling to easily cover the case where the file doesn't exist)
+
+	pkgs, err := packages.Load(
+		&packages.Config{
+			Mode:  packages.LoadFiles,
+			Tests: false,
+		},
+		os.Args[1],
+	)
+	check(err)
+
+	if len(pkgs) != 1 {
+		panic(pkgs)
+	}
+
+	p := pkgs[0]
+
+	var tests []*ast.FuncDecl
+
+	for _, f := range p.GoFiles {
+		s := token.NewFileSet()
+		a, err := parser.ParseFile(s, f, nil, parser.AllErrors)
+		check(err)
+		finder := &platformtestFuncDeclFinder{
+			pkg: p,
+		}
+		ast.Walk(finder, a)
+		tests = append(tests, finder.testFuncs...)
+	}
+
+	sort.Slice(tests, func(i, j int) bool {
+		return strings.Compare(tests[i].Name.Name, tests[j].Name.Name) < 0
+	})
+
+	{
+		casesTemplate := `
+// Code generated by zrepl tooling; DO NOT EDIT.
+
+package tests
+
+var Cases = []Case {
+{{- range . -}}
+	{{ .Name }},
+{{ end -}}
+}
+
+	`
+		t, err := template.New("CaseFunc").Parse(casesTemplate)
+		check(err)
+
+		var buf bytes.Buffer
+		err = t.Execute(&buf, tests)
+		check(err)
+
+		formatted, err := format.Source(buf.Bytes())
+		check(err)
+
+		err = ioutil.WriteFile("generated_cases.go", formatted, 0664)
+		check(err)
+
+	}
+
+}

platformtest/tests/generated_cases.go

@@ -0,0 +1,30 @@
+// Code generated by zrepl tooling; DO NOT EDIT.
+
+package tests
+
+var Cases = []Case{BatchDestroy,
+	CreateReplicationCursor,
+	GetNonexistent,
+	HoldsWork,
+	IdempotentBookmark,
+	IdempotentDestroy,
+	IdempotentHold,
+	ListFilesystemVersionsFilesystemNotExist,
+	ListFilesystemVersionsTypeFilteringAndPrefix,
+	ListFilesystemVersionsUserrefs,
+	ListFilesystemVersionsZeroExistIsNotAnError,
+	ListFilesystemsNoFilter,
+	Pruner2NotReplicated,
+	ReceiveForceIntoEncryptedErr,
+	ReceiveForceRollbackWorksUnencrypted,
+	ReplicationIncrementalCleansUpStaleAbstractionsWithCacheOnSecondReplication,
+	ReplicationIncrementalCleansUpStaleAbstractionsWithoutCacheOnSecondReplication,
+	ReplicationIncrementalIsPossibleIfCommonSnapshotIsDestroyed,
+	ReplicationIsResumableFullSend,
+	ResumableRecvAndTokenHandling,
+	ResumeTokenParsing,
+	SendArgsValidationEncryptedSendOfUnencryptedDatasetForbidden,
+	SendArgsValidationResumeTokenDifferentFilesystemForbidden,
+	SendArgsValidationResumeTokenEncryptionMismatchForbidden,
+	UndestroyableSnapshotParsing,
+}

platformtest/tests/helpers.go

@@ -136,7 +136,7 @@ func makeResumeSituation(ctx *platformtest.Context, src dummySnapshotSituation,
 		return situation
 	}
 
-	limitedCopier := zfs.NewReadCloserCopier(limitio.ReadCloser(copier, src.dummyDataLen/2))
+	limitedCopier := limitio.ReadCloser(copier, src.dummyDataLen/2)
 	defer limitedCopier.Close()
 
 	require.NotNil(ctx, sendArgs.To)

platformtest/tests/holds.go

@@ -0,0 +1,32 @@
+package tests
+
+import (
+	"path"
+
+	"github.com/stretchr/testify/require"
+	"github.com/zrepl/zrepl/platformtest"
+	"github.com/zrepl/zrepl/zfs"
+)
+
+func HoldsWork(ctx *platformtest.Context) {
+	platformtest.Run(ctx, platformtest.PanicErr, ctx.RootDataset, `
+		DESTROYROOT
+		CREATEROOT
+		+  "foo bar"
+		+  "foo bar@snap name"
+	`)
+
+	fs := path.Join(ctx.RootDataset, "foo bar")
+
+	err := zfs.ZFSHold(ctx, fs, fsversion(ctx, fs, "@snap name"), "tag 1")
+	require.NoError(ctx, err)
+
+	err = zfs.ZFSHold(ctx, fs, fsversion(ctx, fs, "@snap name"), "tag 2")
+	require.NoError(ctx, err)
+
+	holds, err := zfs.ZFSHolds(ctx, fs, "snap name")
+	require.NoError(ctx, err)
+	require.Len(ctx, holds, 2)
+	require.Contains(ctx, holds, "tag 1")
+	require.Contains(ctx, holds, "tag 2")
+}

platformtest/tests/idempotentBookmark.go

@@ -3,6 +3,7 @@ package tests
 import (
 	"fmt"
 
+	"github.com/stretchr/testify/assert"
 	"github.com/zrepl/zrepl/platformtest"
 	"github.com/zrepl/zrepl/zfs"
 )
@@ -19,22 +20,23 @@ func IdempotentBookmark(ctx *platformtest.Context) {
 
 	fs := fmt.Sprintf("%s/foo bar", ctx.RootDataset)
 
-	asnap := sendArgVersion(ctx, fs, "@a snap")
-	anotherSnap := sendArgVersion(ctx, fs, "@another snap")
+	asnap := fsversion(ctx, fs, "@a snap")
+	anotherSnap := fsversion(ctx, fs, "@another snap")
 
-	err := zfs.ZFSBookmark(ctx, fs, asnap, "a bookmark")
+	aBookmark, err := zfs.ZFSBookmark(ctx, fs, asnap, "a bookmark")
 	if err != nil {
 		panic(err)
 	}
 
 	// do it again, should be idempotent
-	err = zfs.ZFSBookmark(ctx, fs, asnap, "a bookmark")
+	aBookmarkIdemp, err := zfs.ZFSBookmark(ctx, fs, asnap, "a bookmark")
 	if err != nil {
 		panic(err)
 	}
+	assert.Equal(ctx, aBookmark, aBookmarkIdemp)
 
 	// should fail for another snapshot
-	err = zfs.ZFSBookmark(ctx, fs, anotherSnap, "a bookmark")
+	_, err = zfs.ZFSBookmark(ctx, fs, anotherSnap, "a bookmark")
 	if err == nil {
 		panic(err)
 	}
@@ -48,7 +50,7 @@ func IdempotentBookmark(ctx *platformtest.Context) {
 	}
 
 	// do it again, should fail with special error type
-	err = zfs.ZFSBookmark(ctx, fs, asnap, "a bookmark")
+	_, err = zfs.ZFSBookmark(ctx, fs, asnap, "a bookmark")
 	if err == nil {
 		panic(err)
 	}

platformtest/tests/idempotentDestroy.go

@@ -18,8 +18,8 @@ func IdempotentDestroy(ctx *platformtest.Context) {
 	`)
 
 	fs := fmt.Sprintf("%s/foo bar", ctx.RootDataset)
-	asnap := sendArgVersion(ctx, fs, "@a snap")
-	err := zfs.ZFSBookmark(ctx, fs, asnap, "a bookmark")
+	asnap := fsversion(ctx, fs, "@a snap")
+	_, err := zfs.ZFSBookmark(ctx, fs, asnap, "a bookmark")
 	if err != nil {
 		panic(err)
 	}

platformtest/tests/listFilesystemVersions.go

@@ -43,7 +43,7 @@ func ListFilesystemVersionsTypeFilteringAndPrefix(t *platformtest.Context) {
 	fs := fmt.Sprintf("%s/foo bar", t.RootDataset)
 
 	// no options := all types
-	vs, err := zfs.ZFSListFilesystemVersions(mustDatasetPath(fs), zfs.ListFilesystemVersionsOptions{})
+	vs, err := zfs.ZFSListFilesystemVersions(t, mustDatasetPath(fs), zfs.ListFilesystemVersionsOptions{})
 	require.NoError(t, err)
 	require.Equal(t, []string{
 		"#blup 1", "#bookfoo 1", "#bookfoo 2", "#foo 1", "#foo 2",
@@ -51,21 +51,21 @@ func ListFilesystemVersionsTypeFilteringAndPrefix(t *platformtest.Context) {
 	}, versionRelnamesSorted(vs))
 
 	// just snapshots
-	vs, err = zfs.ZFSListFilesystemVersions(mustDatasetPath(fs), zfs.ListFilesystemVersionsOptions{
+	vs, err = zfs.ZFSListFilesystemVersions(t, mustDatasetPath(fs), zfs.ListFilesystemVersionsOptions{
 		Types: zfs.Snapshots,
 	})
 	require.NoError(t, err)
 	require.Equal(t, []string{"@ foo with leading whitespace", "@blup 1", "@foo 1", "@foo 2"}, versionRelnamesSorted(vs))
 
 	// just bookmarks
-	vs, err = zfs.ZFSListFilesystemVersions(mustDatasetPath(fs), zfs.ListFilesystemVersionsOptions{
+	vs, err = zfs.ZFSListFilesystemVersions(t, mustDatasetPath(fs), zfs.ListFilesystemVersionsOptions{
 		Types: zfs.Bookmarks,
 	})
 	require.NoError(t, err)
 	require.Equal(t, []string{"#blup 1", "#bookfoo 1", "#bookfoo 2", "#foo 1", "#foo 2"}, versionRelnamesSorted(vs))
 
 	// just with prefix foo
-	vs, err = zfs.ZFSListFilesystemVersions(mustDatasetPath(fs), zfs.ListFilesystemVersionsOptions{
+	vs, err = zfs.ZFSListFilesystemVersions(t, mustDatasetPath(fs), zfs.ListFilesystemVersionsOptions{
 		ShortnamePrefix: "foo",
 	})
 	require.NoError(t, err)
@@ -82,12 +82,9 @@ func ListFilesystemVersionsZeroExistIsNotAnError(t *platformtest.Context) {
 
 	fs := fmt.Sprintf("%s/foo bar", t.RootDataset)
 
-	vs, err := zfs.ZFSListFilesystemVersions(mustDatasetPath(fs), zfs.ListFilesystemVersionsOptions{})
+	vs, err := zfs.ZFSListFilesystemVersions(t, mustDatasetPath(fs), zfs.ListFilesystemVersionsOptions{})
 	require.Empty(t, vs)
 	require.NoError(t, err)
-	dsne, ok := err.(*zfs.DatasetDoesNotExist)
-	require.True(t, ok)
-	require.Equal(t, fs, dsne.Path)
 }
 
 func ListFilesystemVersionsFilesystemNotExist(t *platformtest.Context) {
@@ -98,7 +95,7 @@ func ListFilesystemVersionsFilesystemNotExist(t *platformtest.Context) {
 
 	nonexistentFS := fmt.Sprintf("%s/not existent", t.RootDataset)
 
-	vs, err := zfs.ZFSListFilesystemVersions(mustDatasetPath(nonexistentFS), zfs.ListFilesystemVersionsOptions{})
+	vs, err := zfs.ZFSListFilesystemVersions(t, mustDatasetPath(nonexistentFS), zfs.ListFilesystemVersionsOptions{})
 	require.Empty(t, vs)
 	require.Error(t, err)
 	t.Logf("err = %T\n%s", err, err)
@@ -141,7 +138,7 @@ func ListFilesystemVersionsUserrefs(t *platformtest.Context) {
 
 	fs := fmt.Sprintf("%s/foo bar", t.RootDataset)
 
-	vs, err := zfs.ZFSListFilesystemVersions(mustDatasetPath(fs), zfs.ListFilesystemVersionsOptions{})
+	vs, err := zfs.ZFSListFilesystemVersions(t, mustDatasetPath(fs), zfs.ListFilesystemVersionsOptions{})
 	require.NoError(t, err)
 
 	type expectation struct {

platformtest/tests/pruner.go.deact

@@ -0,0 +1,331 @@
+package tests
+
+import (
+	"encoding/json"
+	"fmt"
+	"time"
+
+	"github.com/prometheus/client_golang/prometheus"
+	"github.com/stretchr/testify/require"
+	"github.com/zrepl/zrepl/config"
+	"github.com/zrepl/zrepl/daemon/filters"
+	"github.com/zrepl/zrepl/daemon/pruner"
+	"github.com/zrepl/zrepl/endpoint"
+	"github.com/zrepl/zrepl/platformtest"
+	"github.com/zrepl/zrepl/zfs"
+)
+
+func PrunerNotReplicated(ctx *platformtest.Context) {
+
+	platformtest.Run(ctx, platformtest.PanicErr, ctx.RootDataset, `
+        DESTROYROOT
+        CREATEROOT
+        +  "foo bar"
+        +  "foo bar@1"
+        +  "foo bar@2"
+        +  "foo bar@3"
+        +  "foo bar@4"
+        +  "foo bar@5"
+    `)
+
+	c, err := config.ParseConfigBytes([]byte(fmt.Sprintf(`
+jobs:
+- name: prunetest
+  type: push
+  filesystems: {
+      "%s/foo bar<": true
+  }
+  connect:
+    type: tcp
+    address: 255.255.255.255:255
+  snapshotting:
+    type: manual
+  pruning:
+    keep_sender:
+    - type: not_replicated
+    - type: last_n
+      count: 1
+    keep_receiver:
+    - type: last_n
+      count: 2
+    `, ctx.RootDataset)))
+	require.NoError(ctx, err)
+
+	pushJob := c.Jobs[0].Ret.(*config.PushJob)
+
+	dummyHistVec := prometheus.NewHistogramVec(prometheus.HistogramOpts{
+		Namespace: "foo",
+		Subsystem: "foo",
+		Name:      "foo",
+		Help:      "foo",
+	}, []string{"foo"})
+
+	prunerFactory, err := pruner.NewPrunerFactory(pushJob.Pruning, dummyHistVec)
+	require.NoError(ctx, err)
+
+	senderJid := endpoint.MustMakeJobID("sender-job")
+
+	fsfilter, err := filters.DatasetMapFilterFromConfig(pushJob.Filesystems)
+	require.NoError(ctx, err)
+
+	sender := endpoint.NewSender(endpoint.SenderConfig{
+		FSF: fsfilter,
+		Encrypt: &zfs.NilBool{
+			B: false,
+		},
+		JobID: senderJid,
+	})
+
+	fs := ctx.RootDataset + "/foo bar"
+
+	// create a replication cursor to make pruning work at all
+	_, err = endpoint.CreateReplicationCursor(ctx, fs, fsversion(ctx, fs, "@2"), senderJid)
+	require.NoError(ctx, err)
+
+	p := prunerFactory.BuildSenderPruner(ctx, sender, sender)
+
+	p.Prune()
+
+	report := p.Report()
+
+	reportJSON, err := json.MarshalIndent(report, "", "  ")
+	require.NoError(ctx, err)
+	ctx.Logf("%s\n", string(reportJSON))
+
+	require.Equal(ctx, pruner.Done.String(), report.State)
+	require.Len(ctx, report.Completed, 1)
+	fsReport := report.Completed[0]
+	require.Equal(ctx, fs, fsReport.Filesystem)
+	require.Empty(ctx, fsReport.SkipReason)
+	require.Empty(ctx, fsReport.LastError)
+	require.Len(ctx, fsReport.DestroyList, 1)
+	require.Equal(ctx, fsReport.DestroyList[0], pruner.SnapshotReport{
+		Name:       "1",
+		Replicated: true,
+		Date:       fsReport.DestroyList[0].Date,
+	})
+
+}
+
+func PrunerNoKeepNotReplicatedNoKeepStepHoldConvertsAnyStepHoldToBookmark(ctx *platformtest.Context) {
+
+	platformtest.Run(ctx, platformtest.PanicErr, ctx.RootDataset, `
+        DESTROYROOT
+        CREATEROOT
+        +  "foo bar"
+        +  "foo bar@1"
+        +  "foo bar@2"
+        +  "foo bar@3"
+        +  "foo bar@4"
+        +  "foo bar@5"
+    `)
+
+	c, err := config.ParseConfigBytes([]byte(fmt.Sprintf(`
+jobs:
+- name: prunetest
+  type: push
+  filesystems: {
+      "%s/foo bar<": true
+  }
+  connect:
+    type: tcp
+    address: 255.255.255.255:255
+  snapshotting:
+    type: manual
+  pruning:
+    keep_sender:
+    - type: last_n
+      count: 1
+    keep_receiver:
+    - type: last_n
+      count: 2
+    `, ctx.RootDataset)))
+	require.NoError(ctx, err)
+
+	pushJob := c.Jobs[0].Ret.(*config.PushJob)
+
+	dummyHistVec := prometheus.NewHistogramVec(prometheus.HistogramOpts{
+		Namespace: "foo",
+		Subsystem: "foo",
+		Name:      "foo",
+		Help:      "foo",
+	}, []string{"foo"})
+
+	prunerFactory, err := pruner.NewPrunerFactory(pushJob.Pruning, dummyHistVec)
+	require.NoError(ctx, err)
+
+	senderJid := endpoint.MustMakeJobID("sender-job")
+
+	fsfilter, err := filters.DatasetMapFilterFromConfig(pushJob.Filesystems)
+	require.NoError(ctx, err)
+
+	sender := endpoint.NewSender(endpoint.SenderConfig{
+		FSF: fsfilter,
+		Encrypt: &zfs.NilBool{
+			B: false,
+		},
+        JobID: senderJid,
+	})
+
+	fs := ctx.RootDataset + "/foo bar"
+
+	// create a replication cursor to make pruning work at all
+	_, err = endpoint.CreateReplicationCursor(ctx, fs, fsversion(ctx, fs, "@2"), senderJid)
+	require.NoError(ctx, err)
+
+	// create step holds for the incremental @2->@3
+	endpoint.HoldStep(ctx, fs, fsversion(ctx, fs, "@2"), senderJid)
+	endpoint.HoldStep(ctx, fs, fsversion(ctx, fs, "@3"), senderJid)
+	// create step holds for another job
+	otherJid := endpoint.MustMakeJobID("other-job")
+	endpoint.HoldStep(ctx, fs, fsversion(ctx, fs, "@2"), otherJid)
+	endpoint.HoldStep(ctx, fs, fsversion(ctx, fs, "@3"), otherJid)
+
+	p := prunerFactory.BuildSenderPruner(ctx, sender, sender)
+
+	p.Prune()
+
+	report := p.Report()
+
+	reportJSON, err := json.MarshalIndent(report, "", "  ")
+	require.NoError(ctx, err)
+	ctx.Logf("%s\n", string(reportJSON))
+
+	require.Equal(ctx, pruner.Done.String(), report.State)
+	require.Len(ctx, report.Completed, 1)
+	fsReport := report.Completed[0]
+	require.Equal(ctx, fs, fsReport.Filesystem)
+	require.Empty(ctx, fsReport.SkipReason)
+	require.Empty(ctx, fsReport.LastError)
+	expectDestroyList := []pruner.SnapshotReport{
+		{
+			Name:       "1",
+			Replicated: true,
+		},
+		{
+			Name:       "2",
+			Replicated: true,
+		},
+		{
+			Name:       "3",
+			Replicated: true,
+		},
+		{
+			Name:       "4",
+			Replicated: true,
+		},
+	}
+	for _, d := range fsReport.DestroyList {
+		d.Date = time.Time{}
+	}
+	require.Subset(ctx, fsReport.DestroyList, expectDestroyList)
+}
+
+
+
+func PrunerNoKeepNotReplicatedButKeepStepHold(ctx *platformtest.Context) {
+
+	platformtest.Run(ctx, platformtest.PanicErr, ctx.RootDataset, `
+        DESTROYROOT
+        CREATEROOT
+        +  "foo bar"
+        +  "foo bar@1"
+        +  "foo bar@2"
+        +  "foo bar@3"
+        +  "foo bar@4"
+        +  "foo bar@5"
+    `)
+
+	c, err := config.ParseConfigBytes([]byte(fmt.Sprintf(`
+jobs:
+- name: prunetest
+  type: push
+  filesystems: {
+      "%s/foo bar<": true
+  }
+  connect:
+    type: tcp
+    address: 255.255.255.255:255
+  snapshotting:
+    type: manual
+  pruning:
+    keep_sender:
+    - type: step_holds
+    - type: last_n
+      count: 1
+    keep_receiver:
+    - type: last_n
+      count: 2
+    `, ctx.RootDataset)))
+	require.NoError(ctx, err)
+
+	pushJob := c.Jobs[0].Ret.(*config.PushJob)
+
+	dummyHistVec := prometheus.NewHistogramVec(prometheus.HistogramOpts{
+		Namespace: "foo",
+		Subsystem: "foo",
+		Name:      "foo",
+		Help:      "foo",
+	}, []string{"foo"})
+
+	prunerFactory, err := pruner.NewPrunerFactory(pushJob.Pruning, dummyHistVec)
+	require.NoError(ctx, err)
+
+	senderJid := endpoint.MustMakeJobID("sender-job")
+
+	fsfilter, err := filters.DatasetMapFilterFromConfig(pushJob.Filesystems)
+	require.NoError(ctx, err)
+
+	sender := endpoint.NewSender(endpoint.SenderConfig{
+		FSF: fsfilter,
+		Encrypt: &zfs.NilBool{
+			B: false,
+		},
+        JobID: senderJid,
+	})
+
+	fs := ctx.RootDataset + "/foo bar"
+
+	// create a replication cursor to make pruning work at all
+	_, err = endpoint.CreateReplicationCursor(ctx, fs, fsversion(ctx, fs, "@2"), senderJid)
+	require.NoError(ctx, err)
+
+	// create step holds for the incremental @2->@3
+	endpoint.HoldStep(ctx, fs, fsversion(ctx, fs, "@2"), senderJid)
+	endpoint.HoldStep(ctx, fs, fsversion(ctx, fs, "@3"), senderJid)
+	// create step holds for another job
+	otherJid := endpoint.MustMakeJobID("other-job")
+	endpoint.HoldStep(ctx, fs, fsversion(ctx, fs, "@2"), otherJid)
+	endpoint.HoldStep(ctx, fs, fsversion(ctx, fs, "@3"), otherJid)
+
+	p := prunerFactory.BuildSenderPruner(ctx, sender, sender)
+
+	p.Prune()
+
+	report := p.Report()
+
+	reportJSON, err := json.MarshalIndent(report, "", "  ")
+	require.NoError(ctx, err)
+	ctx.Logf("%s\n", string(reportJSON))
+
+	require.Equal(ctx, pruner.Done.String(), report.State)
+	require.Len(ctx, report.Completed, 1)
+	fsReport := report.Completed[0]
+	require.Equal(ctx, fs, fsReport.Filesystem)
+	require.Empty(ctx, fsReport.SkipReason)
+	require.Empty(ctx, fsReport.LastError)
+	expectDestroyList := []pruner.SnapshotReport{
+		{
+			Name:       "1",
+			Replicated: true,
+		},
+		{
+			Name:       "4",
+			Replicated: true,
+		},
+	}
+	for _, d := range fsReport.DestroyList {
+		d.Date = time.Time{}
+	}
+	require.Subset(ctx, fsReport.DestroyList, expectDestroyList)
+}

platformtest/tests/pruner.v2.go

@@ -0,0 +1,137 @@
+package tests
+
+import (
+	"encoding/json"
+	"fmt"
+	"time"
+
+	"github.com/kr/pretty"
+	"github.com/stretchr/testify/require"
+	"github.com/zrepl/zrepl/config"
+	"github.com/zrepl/zrepl/daemon/filters"
+	"github.com/zrepl/zrepl/daemon/pruner.v2"
+	"github.com/zrepl/zrepl/endpoint"
+	"github.com/zrepl/zrepl/platformtest"
+	"github.com/zrepl/zrepl/pruning"
+	"github.com/zrepl/zrepl/zfs"
+)
+
+func Pruner2NotReplicated(ctx *platformtest.Context) {
+
+	platformtest.Run(ctx, platformtest.PanicErr, ctx.RootDataset, `
+        DESTROYROOT
+        CREATEROOT
+        +  "foo bar"
+        +  "foo bar@1"
+        +  "foo bar@2"
+        +  "foo bar@3"
+        +  "foo bar@4"
+        +  "foo bar@5"
+	`)
+
+	fs := ctx.RootDataset + "/foo bar"
+	senderJid := endpoint.MustMakeJobID("sender-job")
+	otherJid1 := endpoint.MustMakeJobID("other-job-1")
+	otherJid2 := endpoint.MustMakeJobID("other-job-2")
+
+
+	// create step holds for the incremental @2->@3
+	endpoint.HoldStep(ctx, fs, fsversion(ctx, fs, "@2"), senderJid)
+	endpoint.HoldStep(ctx, fs, fsversion(ctx, fs, "@3"), senderJid)
+	// create step holds for other-job-1 @2 -> @3
+	endpoint.HoldStep(ctx, fs, fsversion(ctx, fs, "@2"), otherJid1)
+	endpoint.HoldStep(ctx, fs, fsversion(ctx, fs, "@3"), otherJid1)
+	// create step hold for other-job-2 @1 (will be pruned)
+	endpoint.HoldStep(ctx, fs, fsversion(ctx, fs, "@1"), otherJid2)
+
+	c, err := config.ParseConfigBytes([]byte(fmt.Sprintf(`
+jobs:
+- name: prunetest
+  type: push
+  filesystems: {
+      "%s/foo bar": true
+  }
+  connect:
+    type: tcp
+    address: 255.255.255.255:255
+  snapshotting:
+    type: manual
+  pruning:
+    keep_sender:
+    #- type: not_replicated
+    - type: step_holds
+    - type: last_n
+      count: 1
+    keep_receiver:
+    - type: last_n
+      count: 2
+    `, ctx.RootDataset)))
+	require.NoError(ctx, err)
+
+	pushJob := c.Jobs[0].Ret.(*config.PushJob)
+
+	require.NoError(ctx, err)
+
+	fsfilter, err := filters.DatasetMapFilterFromConfig(pushJob.Filesystems)
+	require.NoError(ctx, err)
+
+	matchedFilesystems, err := zfs.ZFSListMapping(ctx, fsfilter)
+	ctx.Logf("%s", pretty.Sprint(matchedFilesystems))
+	require.NoError(ctx, err)
+	require.Len(ctx, matchedFilesystems, 1)
+
+	sideSender := pruner.NewSideSender(senderJid)
+
+	keepRules, err := pruning.RulesFromConfig(senderJid, pushJob.Pruning.KeepSender)
+	require.NoError(ctx, err)
+	p := pruner.NewPruner(fsfilter, senderJid, sideSender, keepRules)
+
+	runDone := make(chan *pruner.Report)
+	go func() {
+		runDone <- p.Run(ctx)
+	}()
+
+	var report *pruner.Report
+	// concurrency stress
+out:
+	for {
+		select {
+		case <-time.After(10 * time.Millisecond):
+			p.Report(ctx)
+		case report = <-runDone:
+			break out
+		}
+	}
+	ctx.Logf("%s\n", pretty.Sprint(report))
+
+	reportJSON, err := json.MarshalIndent(report, "", "  ")
+	require.NoError(ctx, err)
+	ctx.Logf("%s\n", string(reportJSON))
+
+	ctx.FailNow()
+	// fs := ctx.RootDataset + "/foo bar"
+
+	// // create a replication cursor to make pruning work at all
+	// _, err = endpoint.CreateReplicationCursor(ctx, fs, fsversion(ctx, fs, "@2"), senderJid)
+	// require.NoError(ctx, err)
+
+	// p := prunerFactory.BuildSenderPruner(ctx, sender, sender)
+
+	// p.Prune()
+
+	// report := p.Report()
+
+	// require.Equal(ctx, pruner.Done.String(), report.State)
+	// require.Len(ctx, report.Completed, 1)
+	// fsReport := report.Completed[0]
+	// require.Equal(ctx, fs, fsReport.Filesystem)
+	// require.Empty(ctx, fsReport.SkipReason)
+	// require.Empty(ctx, fsReport.LastError)
+	// require.Len(ctx, fsReport.DestroyList, 1)
+	// require.Equal(ctx, fsReport.DestroyList[0], pruner.SnapshotReport{
+	// 	Name:       "1",
+	// 	Replicated: true,
+	// 	Date:       fsReport.DestroyList[0].Date,
+	// })
+
+}

platformtest/tests/recvForceIntoEncryptedErr.go

@@ -0,0 +1,45 @@
+package tests
+
+import (
+	"fmt"
+
+	"github.com/stretchr/testify/require"
+	"github.com/zrepl/zrepl/platformtest"
+	"github.com/zrepl/zrepl/zfs"
+)
+
+func ReceiveForceIntoEncryptedErr(ctx *platformtest.Context) {
+	platformtest.Run(ctx, platformtest.PanicErr, ctx.RootDataset, `
+		DESTROYROOT
+		CREATEROOT
+		+  "foo bar" encrypted
+		+  "sender" encrypted
+		+  "sender@1"
+	`)
+
+	rfs := fmt.Sprintf("%s/foo bar", ctx.RootDataset)
+	sfs := fmt.Sprintf("%s/sender", ctx.RootDataset)
+	sfsSnap1 := sendArgVersion(ctx, sfs, "@1")
+
+	sendArgs, err := zfs.ZFSSendArgsUnvalidated{
+		FS:          sfs,
+		Encrypted:   &zfs.NilBool{B: false},
+		From:        nil,
+		To:          &sfsSnap1,
+		ResumeToken: "",
+	}.Validate(ctx)
+	require.NoError(ctx, err)
+
+	sendStream, err := zfs.ZFSSend(ctx, sendArgs)
+	require.NoError(ctx, err)
+
+	recvOpts := zfs.RecvOptions{
+		RollbackAndForceRecv: true,
+		SavePartialRecvState: false,
+	}
+	err = zfs.ZFSRecv(ctx, rfs, &zfs.ZFSSendArgVersion{RelName: "@1", GUID: sfsSnap1.GUID}, sendStream, recvOpts)
+	require.Error(ctx, err)
+	re, ok := err.(*zfs.RecvDestroyOrOverwriteEncryptedErr)
+	require.True(ctx, ok)
+	require.Contains(ctx, re.Error(), "zfs receive -F cannot be used to destroy an encrypted filesystem or overwrite an unencrypted one with an encrypted on")
+}

platformtest/tests/recvRollback.go

@@ -0,0 +1,54 @@
+package tests
+
+import (
+	"fmt"
+
+	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
+	"github.com/zrepl/zrepl/platformtest"
+	"github.com/zrepl/zrepl/zfs"
+)
+
+func ReceiveForceRollbackWorksUnencrypted(ctx *platformtest.Context) {
+	platformtest.Run(ctx, platformtest.PanicErr, ctx.RootDataset, `
+		DESTROYROOT
+		CREATEROOT
+		+  "foo bar"
+		+  "foo bar@a snap"
+		+  "foo bar@another snap"
+		+  "foo bar@snap3"
+		+  "sender"
+		+  "sender@1"
+	`)
+
+	rfs := fmt.Sprintf("%s/foo bar", ctx.RootDataset)
+	sfs := fmt.Sprintf("%s/sender", ctx.RootDataset)
+	sfsSnap1 := sendArgVersion(ctx, sfs, "@1")
+
+	sendArgs, err := zfs.ZFSSendArgsUnvalidated{
+		FS:          sfs,
+		Encrypted:   &zfs.NilBool{B: false},
+		From:        nil,
+		To:          &sfsSnap1,
+		ResumeToken: "",
+	}.Validate(ctx)
+	require.NoError(ctx, err)
+
+	sendStream, err := zfs.ZFSSend(ctx, sendArgs)
+	require.NoError(ctx, err)
+
+	recvOpts := zfs.RecvOptions{
+		RollbackAndForceRecv: true,
+		SavePartialRecvState: false,
+	}
+	err = zfs.ZFSRecv(ctx, rfs, &zfs.ZFSSendArgVersion{RelName: "@1", GUID: sfsSnap1.GUID}, sendStream, recvOpts)
+	require.NoError(ctx, err)
+
+	// assert exists on receiver
+	rfsSnap1 := fsversion(ctx, rfs, "@1")
+	// assert it's the only one (rollback and force-recv should be blowing away the other filesystems)
+	rfsVersions, err := zfs.ZFSListFilesystemVersions(ctx, mustDatasetPath(rfs), zfs.ListFilesystemVersionsOptions{})
+	require.NoError(ctx, err)
+	assert.Len(ctx, rfsVersions, 1)
+	assert.Equal(ctx, rfsVersions[0], rfsSnap1)
+}

platformtest/tests/replication.go

@@ -0,0 +1,396 @@
+package tests
+
+import (
+	"context"
+	"fmt"
+	"io"
+	"path"
+
+	"github.com/kr/pretty"
+	"github.com/stretchr/testify/require"
+	"github.com/zrepl/zrepl/daemon/filters"
+	"github.com/zrepl/zrepl/endpoint"
+	"github.com/zrepl/zrepl/platformtest"
+	"github.com/zrepl/zrepl/replication"
+	"github.com/zrepl/zrepl/replication/logic"
+	"github.com/zrepl/zrepl/replication/logic/pdu"
+	"github.com/zrepl/zrepl/replication/report"
+	"github.com/zrepl/zrepl/util/limitio"
+	"github.com/zrepl/zrepl/zfs"
+)
+
+// mimics the replication invocations of an active-side job
+// for a single sender-receiver filesystem pair
+//
+// each invocation of method Do results in the construction
+// of a new sender and receiver instance and one blocking invocation
+// of the replication engine without encryption
+type replicationInvocation struct {
+	sjid, rjid      endpoint.JobID
+	sfs             string
+	rfsRoot         string
+	interceptSender func(e *endpoint.Sender) logic.Sender
+}
+
+func (i replicationInvocation) Do(ctx *platformtest.Context) *report.Report {
+
+	if i.interceptSender == nil {
+		i.interceptSender = func(e *endpoint.Sender) logic.Sender { return e }
+	}
+
+	sfilter := filters.NewDatasetMapFilter(1, true)
+	err := sfilter.Add(i.sfs, "ok")
+	require.NoError(ctx, err)
+	sender := i.interceptSender(endpoint.NewSender(endpoint.SenderConfig{
+		FSF:     sfilter.AsFilter(),
+		Encrypt: &zfs.NilBool{B: false},
+		JobID:   i.sjid,
+	}))
+	receiver := endpoint.NewReceiver(endpoint.ReceiverConfig{
+		JobID:                      i.rjid,
+		AppendClientIdentity:       false,
+		RootWithoutClientComponent: mustDatasetPath(i.rfsRoot),
+		UpdateLastReceivedHold:     true,
+	})
+	plannerPolicy := logic.PlannerPolicy{
+		EncryptedSend: logic.TriFromBool(false),
+	}
+
+	report, wait := replication.Do(
+		ctx,
+		logic.NewPlanner(nil, nil, sender, receiver, plannerPolicy),
+	)
+	wait(true)
+	return report()
+}
+
+func (i replicationInvocation) ReceiveSideFilesystem() string {
+	return path.Join(i.rfsRoot, i.sfs)
+}
+
+func ReplicationIncrementalIsPossibleIfCommonSnapshotIsDestroyed(ctx *platformtest.Context) {
+
+	platformtest.Run(ctx, platformtest.PanicErr, ctx.RootDataset, `
+		CREATEROOT
+		+  "sender"
+		+  "sender@1"
+		+  "receiver"
+		R  zfs create -p "${ROOTDS}/receiver/${ROOTDS}"
+	`)
+
+	sjid := endpoint.MustMakeJobID("sender-job")
+	rjid := endpoint.MustMakeJobID("receiver-job")
+
+	sfs := ctx.RootDataset + "/sender"
+	rfsRoot := ctx.RootDataset + "/receiver"
+	snap1 := fsversion(ctx, sfs, "@1")
+
+	rep := replicationInvocation{
+		sjid:    sjid,
+		rjid:    rjid,
+		sfs:     sfs,
+		rfsRoot: rfsRoot,
+	}
+	rfs := rep.ReceiveSideFilesystem()
+
+	// first replication
+	report := rep.Do(ctx)
+	ctx.Logf("\n%s", pretty.Sprint(report))
+
+	// assert @1 exists on receiver
+	_ = fsversion(ctx, rfs, "@1")
+
+	// cut off the common base between sender and receiver
+	// (replication engine guarantees resumability through bookmarks)
+	err := zfs.ZFSDestroy(ctx, snap1.FullPath(sfs))
+	require.NoError(ctx, err)
+
+	// assert that the replication cursor has been created
+	snap1CursorName, err := endpoint.ReplicationCursorBookmarkName(sfs, snap1.Guid, sjid)
+	require.NoError(ctx, err)
+	snap1CursorInfo, err := zfs.ZFSGetFilesystemVersion(ctx, sfs+"#"+snap1CursorName)
+	require.NoError(ctx, err)
+	require.True(ctx, snap1CursorInfo.IsBookmark())
+
+	// second replication of a new snapshot, should use the cursor
+	mustSnapshot(ctx, sfs+"@2")
+	report = rep.Do(ctx)
+	ctx.Logf("\n%s", pretty.Sprint(report))
+	_ = fsversion(ctx, rfs, "@2")
+
+}
+
+func ReplicationIncrementalCleansUpStaleAbstractionsWithCacheOnSecondReplication(ctx *platformtest.Context) {
+	implReplicationIncrementalCleansUpStaleAbstractions(ctx, true)
+}
+
+func ReplicationIncrementalCleansUpStaleAbstractionsWithoutCacheOnSecondReplication(ctx *platformtest.Context) {
+	implReplicationIncrementalCleansUpStaleAbstractions(ctx, false)
+}
+
+func implReplicationIncrementalCleansUpStaleAbstractions(ctx *platformtest.Context, invalidateCacheBeforeSecondReplication bool) {
+
+	platformtest.Run(ctx, platformtest.PanicErr, ctx.RootDataset, `
+		CREATEROOT
+		+  "sender"
+		+  "sender@1"
+		+  "sender@2"
+		+  "sender#2" "sender@2"
+		+  "sender@3"
+		+  "receiver"
+		R  zfs create -p "${ROOTDS}/receiver/${ROOTDS}"
+	`)
+
+	sjid := endpoint.MustMakeJobID("sender-job")
+	ojid := endpoint.MustMakeJobID("other-job")
+	rjid := endpoint.MustMakeJobID("receiver-job")
+
+	sfs := ctx.RootDataset + "/sender"
+	rfsRoot := ctx.RootDataset + "/receiver"
+
+	rep := replicationInvocation{
+		sjid:    sjid,
+		rjid:    rjid,
+		sfs:     sfs,
+		rfsRoot: rfsRoot,
+	}
+	rfs := rep.ReceiveSideFilesystem()
+
+	// first replication
+	report := rep.Do(ctx)
+	ctx.Logf("\n%s", pretty.Sprint(report))
+
+	// assert most recent send-side version @3 exists on receiver (=replication succeeded)
+	rSnap3 := fsversion(ctx, rfs, "@3")
+	// assert the source-side versions not managed by zrepl still exist
+	snap1 := fsversion(ctx, sfs, "@1")
+	snap2 := fsversion(ctx, sfs, "@2")
+	_ = fsversion(ctx, sfs, "#2") // non-replicationc-cursor bookmarks should not be affected
+	snap3 := fsversion(ctx, sfs, "@3")
+	// assert a replication cursor is in place
+	snap3CursorName, err := endpoint.ReplicationCursorBookmarkName(sfs, snap3.Guid, sjid)
+	require.NoError(ctx, err)
+	_ = fsversion(ctx, sfs, "#"+snap3CursorName)
+	// assert a last-received hold is in place
+	expectRjidHoldTag, err := endpoint.LastReceivedHoldTag(rjid)
+	require.NoError(ctx, err)
+	holds, err := zfs.ZFSHolds(ctx, rfs, rSnap3.Name)
+	require.NoError(ctx, err)
+	require.Contains(ctx, holds, expectRjidHoldTag)
+
+	// create artifical stale replication cursors
+	createArtificalStaleAbstractions := func(jobId endpoint.JobID) []endpoint.Abstraction {
+		snap2Cursor, err := endpoint.CreateReplicationCursor(ctx, sfs, snap2, jobId) // no shadow
+		require.NoError(ctx, err)
+		// create artifical stale step holds jobId
+		snap1Hold, err := endpoint.HoldStep(ctx, sfs, snap1, jobId) // no shadow
+		require.NoError(ctx, err)
+		snap2Hold, err := endpoint.HoldStep(ctx, sfs, snap2, jobId) // no shadow
+		require.NoError(ctx, err)
+		return []endpoint.Abstraction{snap2Cursor, snap1Hold, snap2Hold}
+	}
+	createArtificalStaleAbstractions(sjid)
+	ojidSendAbstractions := createArtificalStaleAbstractions(ojid)
+
+	snap3ojidLastReceivedHold, err := endpoint.CreateLastReceivedHold(ctx, rfs, fsversion(ctx, rfs, "@3"), ojid)
+	require.NoError(ctx, err)
+	require.True(ctx, zfs.FilesystemVersionEqualIdentity(fsversion(ctx, rfs, "@3"), snap3ojidLastReceivedHold.GetFilesystemVersion()))
+
+	// take another 2 snapshots
+	mustSnapshot(ctx, sfs+"@4")
+	mustSnapshot(ctx, sfs+"@5")
+	snap5 := fsversion(ctx, sfs, "@5")
+
+	if invalidateCacheBeforeSecondReplication {
+		endpoint.SendAbstractionsCacheInvalidate(sfs)
+	}
+
+	// do another replication
+	// - ojid's abstractions should not be affected on either side
+	// - stale abstractions of sjid and rjid should be cleaned up
+	// - 1 replication cursors and 1 last-received hold should be present
+
+	checkOjidAbstractionsExist := func() {
+		var expectedOjidAbstractions []endpoint.Abstraction
+		expectedOjidAbstractions = append(expectedOjidAbstractions, ojidSendAbstractions...)
+		expectedOjidAbstractions = append(expectedOjidAbstractions, snap3ojidLastReceivedHold)
+
+		sfsAndRfsFilter := filters.NewDatasetMapFilter(2, true)
+		require.NoError(ctx, sfsAndRfsFilter.Add(sfs, "ok"))
+		require.NoError(ctx, sfsAndRfsFilter.Add(rfs, "ok"))
+		rAbs, rAbsErrs, err := endpoint.ListAbstractions(ctx, endpoint.ListZFSHoldsAndBookmarksQuery{
+			FS:          endpoint.ListZFSHoldsAndBookmarksQueryFilesystemFilter{Filter: sfsAndRfsFilter},
+			JobID:       &ojid,
+			What:        endpoint.AbstractionTypesAll,
+			Concurrency: 1,
+		})
+		require.NoError(ctx, err)
+		require.Len(ctx, rAbsErrs, 0)
+		ctx.Logf("rAbs=%s", rAbs)
+		ctx.Logf("expectedOjidAbstractions=%s", expectedOjidAbstractions)
+		require.Equal(ctx, len(expectedOjidAbstractions), len(rAbs))
+		for _, ea := range expectedOjidAbstractions {
+			ctx.Logf("looking for %s %#v", ea, ea.GetFilesystemVersion())
+			found := false
+			for _, a := range rAbs {
+				eq := endpoint.AbstractionEquals(ea, a)
+				ctx.Logf("comp=%v for %s %#v", eq, a, a.GetFilesystemVersion())
+				found = found || eq
+			}
+			require.True(ctx, found, "%s", ea)
+		}
+	}
+	checkOjidAbstractionsExist()
+
+	report = rep.Do(ctx)
+	ctx.Logf("\n%s", pretty.Sprint(report))
+
+	checkOjidAbstractionsExist()
+
+	_ = fsversion(ctx, sfs, "@1")
+	_ = fsversion(ctx, sfs, "@2")
+	_ = fsversion(ctx, sfs, "#2")
+	_ = fsversion(ctx, sfs, "@3")
+	_ = fsversion(ctx, sfs, "@4")
+	_ = fsversion(ctx, sfs, "@5")
+
+	_ = fsversion(ctx, rfs, "@3")
+	_ = fsversion(ctx, rfs, "@4")
+	_ = fsversion(ctx, rfs, "@5")
+
+	// check bookmark situation
+	{
+		sBms, err := zfs.ZFSListFilesystemVersions(ctx, mustDatasetPath(sfs), zfs.ListFilesystemVersionsOptions{
+			Types: zfs.Bookmarks,
+		})
+		ctx.Logf("sbms=%s", sBms)
+		require.NoError(ctx, err)
+
+		snap5SjidCursorName, err := endpoint.ReplicationCursorBookmarkName(sfs, snap5.Guid, sjid)
+		require.NoError(ctx, err)
+		snap2SjidCursorName, err := endpoint.ReplicationCursorBookmarkName(sfs, snap2.Guid, sjid)
+		require.NoError(ctx, err)
+		snap2OjidCursorName, err := endpoint.ReplicationCursorBookmarkName(sfs, snap2.Guid, ojid)
+		require.NoError(ctx, err)
+		var bmNames []string
+		for _, bm := range sBms {
+			bmNames = append(bmNames, bm.Name)
+		}
+
+		if invalidateCacheBeforeSecondReplication {
+			require.Len(ctx, sBms, 3)
+			require.Contains(ctx, bmNames, snap5SjidCursorName)
+			require.Contains(ctx, bmNames, snap2OjidCursorName)
+			require.Contains(ctx, bmNames, "2")
+		} else {
+			require.Len(ctx, sBms, 4)
+			require.Contains(ctx, bmNames, snap5SjidCursorName)
+			require.Contains(ctx, bmNames, snap2SjidCursorName)
+			require.Contains(ctx, bmNames, snap2OjidCursorName)
+			require.Contains(ctx, bmNames, "2")
+		}
+	}
+
+	// check last-received hold moved
+	{
+		rAbs, rAbsErrs, err := endpoint.ListAbstractions(ctx, endpoint.ListZFSHoldsAndBookmarksQuery{
+			FS:          endpoint.ListZFSHoldsAndBookmarksQueryFilesystemFilter{FS: &rfs},
+			JobID:       &rjid,
+			What:        endpoint.AbstractionTypesAll,
+			Concurrency: 1,
+		})
+		require.NoError(ctx, err)
+		require.Len(ctx, rAbsErrs, 0)
+		require.Len(ctx, rAbs, 1)
+		require.Equal(ctx, rAbs[0].GetType(), endpoint.AbstractionLastReceivedHold)
+		require.Equal(ctx, *rAbs[0].GetJobID(), rjid)
+		require.Equal(ctx, rAbs[0].GetFilesystemVersion().GetGuid(), snap5.GetGuid())
+	}
+
+}
+
+type PartialSender struct {
+	*endpoint.Sender
+	failAfterByteCount int64
+}
+
+var _ logic.Sender = (*PartialSender)(nil)
+
+func (s *PartialSender) Send(ctx context.Context, r *pdu.SendReq) (r1 *pdu.SendRes, r2 io.ReadCloser, r3 error) {
+	r1, r2, r3 = s.Sender.Send(ctx, r)
+	r2 = limitio.ReadCloser(r2, s.failAfterByteCount)
+	return r1, r2, r3
+}
+
+func ReplicationIsResumableFullSend(ctx *platformtest.Context) {
+
+	platformtest.Run(ctx, platformtest.PanicErr, ctx.RootDataset, `
+		CREATEROOT
+		+  "sender"
+		+  "receiver"
+		R  zfs create -p "${ROOTDS}/receiver/${ROOTDS}"
+	`)
+
+	sjid := endpoint.MustMakeJobID("sender-job")
+	rjid := endpoint.MustMakeJobID("receiver-job")
+
+	sfs := ctx.RootDataset + "/sender"
+	rfsRoot := ctx.RootDataset + "/receiver"
+
+	sfsmp, err := zfs.ZFSGetMountpoint(ctx, sfs)
+	require.NoError(ctx, err)
+	require.True(ctx, sfsmp.Mounted)
+
+	writeDummyData(path.Join(sfsmp.Mountpoint, "dummy.data"), 1<<22)
+	mustSnapshot(ctx, sfs+"@1")
+	snap1 := fsversion(ctx, sfs, "@1")
+
+	rep := replicationInvocation{
+		sjid:    sjid,
+		rjid:    rjid,
+		sfs:     sfs,
+		rfsRoot: rfsRoot,
+		interceptSender: func(e *endpoint.Sender) logic.Sender {
+			return &PartialSender{Sender: e, failAfterByteCount: 1 << 20}
+		},
+	}
+	rfs := rep.ReceiveSideFilesystem()
+
+	for i := 2; i < 10; i++ {
+		report := rep.Do(ctx)
+		ctx.Logf("\n%s", pretty.Sprint(report))
+
+		// always attempt to destroy the incremental source
+		err := zfs.ZFSDestroy(ctx, snap1.FullPath(sfs))
+		if i < 4 {
+			// we configured the PartialSender to fail after 1<<20 bytes
+			// and we wrote dummy data 1<<22 bytes, thus at least
+			// for the first 4 times this should not be possible
+			// due to step holds
+			require.Error(ctx, err)
+			require.Contains(ctx, err.Error(), "dataset is busy")
+		}
+
+		// and create some additional snapshots that could
+		// confuse a naive implementation that doesn't take into
+		// account resume state when planning replication
+		if i == 2 || i == 3 {
+			// no significant size to avoid making this test run longer than necessary
+			mustSnapshot(ctx, fmt.Sprintf("%s@%d", sfs, i))
+		}
+
+		require.Len(ctx, report.Attempts, 1)
+		require.Nil(ctx, report.Attempts[0].PlanError)
+		require.Len(ctx, report.Attempts[0].Filesystems, 1)
+		if len(report.Attempts[0].Filesystems[0].Steps) == 0 {
+			break
+		}
+	}
+
+	// make sure all the filesystem versions we created
+	// were replicated by the replication loop
+	_ = fsversion(ctx, rfs, "@1")
+	_ = fsversion(ctx, rfs, "@2")
+	_ = fsversion(ctx, rfs, "@3")
+
+}

platformtest/tests/replicationCursor.go

@@ -11,7 +11,7 @@ import (
 	"github.com/zrepl/zrepl/zfs"
 )
 
-func ReplicationCursor(ctx *platformtest.Context) {
+func CreateReplicationCursor(ctx *platformtest.Context) {
 
 	platformtest.Run(ctx, platformtest.PanicErr, ctx.RootDataset, `
 		CREATEROOT
@@ -19,9 +19,9 @@ func ReplicationCursor(ctx *platformtest.Context) {
 		+  "foo bar@1 with space"
 		R  zfs bookmark "${ROOTDS}/foo bar@1 with space" "${ROOTDS}/foo bar#1 with space"
 		+  "foo bar@2 with space"
-		R  zfs bookmark "${ROOTDS}/foo bar@1 with space" "${ROOTDS}/foo bar#2 with space"
+		R  zfs bookmark "${ROOTDS}/foo bar@2 with space" "${ROOTDS}/foo bar#2 with space"
 		+  "foo bar@3 with space"
-		R  zfs bookmark "${ROOTDS}/foo bar@1 with space" "${ROOTDS}/foo bar#3 with space"
+		R  zfs bookmark "${ROOTDS}/foo bar@3 with space" "${ROOTDS}/foo bar#3 with space"
 	`)
 
 	jobid := endpoint.MustMakeJobID("zreplplatformtest")
@@ -32,13 +32,29 @@ func ReplicationCursor(ctx *platformtest.Context) {
 	}
 
 	fs := ds.ToString()
-	snap := fsversion(ctx, fs, "@1 with space")
 
-	destroyed, err := endpoint.MoveReplicationCursor(ctx, fs, &snap, jobid)
-	if err != nil {
-		panic(err)
+	checkCreateCursor := func(createErr error, c endpoint.Abstraction, references zfs.FilesystemVersion) {
+		assert.NoError(ctx, createErr)
+		expectName, err := endpoint.ReplicationCursorBookmarkName(fs, references.Guid, jobid)
+		assert.NoError(ctx, err)
+		require.Equal(ctx, expectName, c.GetFilesystemVersion().Name)
 	}
-	assert.Empty(ctx, destroyed)
+
+	snap := fsversion(ctx, fs, "@1 with space")
+	book := fsversion(ctx, fs, "#1 with space")
+
+	// create first cursor
+	cursorOfSnap, err := endpoint.CreateReplicationCursor(ctx, fs, snap, jobid)
+	checkCreateCursor(err, cursorOfSnap, snap)
+	// check CreateReplicationCursor is idempotent (for snapshot target)
+	cursorOfSnapIdemp, err := endpoint.CreateReplicationCursor(ctx, fs, snap, jobid)
+	checkCreateCursor(err, cursorOfSnap, snap)
+	// ... for target = non-cursor bookmark
+	_, err = endpoint.CreateReplicationCursor(ctx, fs, book, jobid)
+	assert.Equal(ctx, zfs.ErrBookmarkCloningNotSupported, err)
+	// ... for target = replication cursor bookmark to be created
+	cursorOfCursor, err := endpoint.CreateReplicationCursor(ctx, fs, cursorOfSnapIdemp.GetFilesystemVersion(), jobid)
+	checkCreateCursor(err, cursorOfCursor, cursorOfSnap.GetFilesystemVersion())
 
 	snapProps, err := zfs.ZFSGetFilesystemVersion(ctx, snap.FullPath(fs))
 	if err != nil {
@@ -56,14 +72,4 @@ func ReplicationCursor(ctx *platformtest.Context) {
 		panic(fmt.Sprintf("guids do not match: %v != %v", bm.Guid, snapProps.Guid))
 	}
 
-	// try moving
-	cursor1BookmarkName, err := endpoint.ReplicationCursorBookmarkName(fs, snap.Guid, jobid)
-	require.NoError(ctx, err)
-
-	snap2 := fsversion(ctx, fs, "@2 with space")
-	destroyed, err = endpoint.MoveReplicationCursor(ctx, fs, &snap2, jobid)
-	require.NoError(ctx, err)
-	require.Equal(ctx, 1, len(destroyed))
-	require.Equal(ctx, endpoint.AbstractionReplicationCursorBookmarkV2, destroyed[0].GetType())
-	require.Equal(ctx, cursor1BookmarkName, destroyed[0].GetName())
 }