Compare commits

..

6 Commits

Author SHA1 Message Date
Christian Schwarz 0ea8f96db6 replication/driver: fs.debug() helper that automatically prefixes with fs name 2020-04-07 23:43:06 +02:00
Christian Schwarz 3cc0acd7ec replication/driver: rename receiver variable (fs *fs) to (f *fs) 2020-04-07 23:41:28 +02:00
Christian Schwarz 632acb6305 replication/driver: envconst for experimental parallel replication
refs #140
refs #302
2020-04-07 23:35:42 +02:00
Christian Schwarz 4376346d2c endpoint: log %#v recv options 2020-04-07 23:33:59 +02:00
Christian Schwarz 5a00b35d6f build: Makefile: GO_EXTRA_BUILDFLAGS 2020-04-07 23:33:42 +02:00
Christian Schwarz 98def6e940 endpoint: refactor, fix stale holds on initial replication failure, holds release subcmds, more efficient ZFS queries
The motivation for this recatoring are based on two independent issues:

- @JMoVS found that the changes merged as part of #259 slowed his OS X
  based installation down significantly.
  Analysis of the zfs command logging introduced in #296 showed that
  `zfs holds` took most of the execution time, and they pointed out
  that not all of those `zfs holds` invocations were actually necessary.
  I.e.: zrepl was inefficient about retrieving information from ZFS.

- @InsanePrawn found that failures on initial replication would lead
  to step holds accumulating on the sending side, i.e. they would never
  be cleaned up in the HintMostRecentCommonAncestor RPC handler.
  That was because we only sent that RPC if there was a most recent
  common ancestor detected during replication planning.
  @InsanePrawn prototyped an implementation of a `zrepl holds release`
  command to mitigate the situation.
  As part of that development work and back-and-forth with @problame,
  it became evident that the abstractions that #259 built on top of
  zfs in package endpoint (step holds, replication cursor,
  last-received-hold), were not well-represented for re-use in the
  `zrepl holds release` subocommand.

This commit refactors package endpoint to address both of these issues:

- endpoint abstractions now share an interface `Abstraction` that, among
  other things, provides a uniform `Destroy()` method.
  However, that method should not be destroyed directly but instead
  the package-level `BatchDestroy` function should be used in order
  to allow for a migration to zfs channel programs in the future.

- endpoint now has a query facitilty (`ListAbstractions`) which is
  used to find on-disk
    - step holds and bookmarks
    - replication cursors (v1, v2)
    - last-received-holds
  By describing the query in a struct, we can centralized the retrieval
  of information via the ZFS CLI and only have to be clever once.
  We are "clever" in the following ways:
  - When asking for hold-based abstractions, we only run `zfs holds` on
    snapshot that have `userrefs` > 0
    - To support this functionality, add field `UserRefs` to zfs.FilesystemVersion
      and retrieve it anywhere we retrieve zfs.FilesystemVersion from ZFS.
  - When asking only for bookmark-based abstractions, we only run
    `zfs list -t bookmark`, not with snapshots.
  - Currently unused (except for CLI) per-filesystem concurrent lookup
  - Option to only include abstractions with CreateTXG in a specified range

- refactor `endpoint`'s various ZFS info  retrieval methods to use
  `ListAbstractions`

- change the `zrepl holds list` command to consume endpoint.ListAbstractions

- Add a `ListStale` method which, given a query template,
  lists stale holds and bookmarks.
  - it uses replication cursor has different modes
- the new `zrepl holds release-{all,stale}` commands can be used
  to remove abstractions of package endpoint

- Adjust HintMostRecentCommonAncestor RPC for stale-holds cleanup:
    - send it also if no most recent common ancestor exists between sender and receiver
    - have the sender clean up its abstractions when it receives the RPC
      with no most recent common ancestor, using `ListStale`
    - Due to changed semantics, bump the protocol version.

- Adjust HintMostRecentCommonAncestor RPC for performance problems
  encountered by @JMoVS
    - by default, per (job,fs)-combination, only consider cleaning
      step holds in the createtxg range
      `[last replication cursor,conservatively-estimated-receive-side-version)`
    - this behavior ensures resumability at cost proportional to the
      time that replication was donw
    - however, as explained in a comment, we might leak holds if
      the zrepl daemon stops running
    - that  trade-off is acceptable because in the presumably rare
      this might happen the user has two tools at their hand:
    - Tool 1: run `zrepl holds release-stale`
    - Tool 2: use env var `ZREPL_ENDPOINT_SENDER_HINT_MOST_RECENT_STEP_HOLD_CLEANUP_MODE`
      to adjust the lower bound of the createtxg range (search for it in the code).
      The env var can also be used to disable hold-cleanup on the
      send-side entirely.

supersedes closes #293
supersedes closes #282
fixes #280
fixes #278

Additionaly, we fixed a couple of bugs:

- zfs: fix half-nil error reporting of dataset-does-not-exist for ZFSListChan and ZFSBookmark

- endpoint: Sender's `HintMostRecentCommonAncestor` handler would not
  check whether access to the specified filesystem was allowed.
2020-04-07 18:57:55 +02:00
16 changed files with 131 additions and 212 deletions
+1 -11
View File
@@ -92,21 +92,11 @@ jobs:
REPO="zrepl/zrepl"
COMMIT="${CIRCLE_SHA1}"
JOB_NAME="${CIRCLE_JOB}"
curl "https://api.github.com/repos/$REPO/statuses/$COMMIT" \
curl "https://api.github.com/repos/$REPO/statuses/$COMMIT?access_token=$GITHUB_COMMIT_STATUS_TOKEN" \
-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:
+5 -9
View File
@@ -23,7 +23,8 @@ GOHOSTARCH ?= $(shell bash -c 'source <($(GO) env) && echo "$$GOHOSTARCH"')
GO_ENV_VARS := GO111MODULE=on
GO_LDFLAGS := "-X github.com/zrepl/zrepl/version.zreplVersion=$(_ZREPL_VERSION)"
GO_MOD_READONLY := -mod=readonly
GO_BUILDFLAGS := $(GO_MOD_READONLY)
GO_EXTRA_BUILDFLAGS :=
GO_BUILDFLAGS := $(GO_MOD_READONLY) $(GO_EXTRA_BUILDFLAGS)
GO_BUILD := $(GO_ENV_VARS) $(GO) build $(GO_BUILDFLAGS) -ldflags $(GO_LDFLAGS)
GOLANGCI_LINT := golangci-lint
ifneq ($(GOARM),)
@@ -65,7 +66,6 @@ wrapup-and-checksum:
-acf $(NOARCH_TARBALL) \
$(ARTIFACTDIR)/docs/html \
$(ARTIFACTDIR)/bash_completion \
$(ARTIFACTDIR)/_zrepl.zsh_completion \
$(ARTIFACTDIR)/go_env.txt \
dist \
config/samples
@@ -160,7 +160,7 @@ platformtest: # do not track dependency on platformtest-bin to allow build of pl
$(ZREPL_PLATFORMTEST_ARGS)
##################### NOARCH #####################
.PHONY: noarch $(ARTIFACTDIR)/bash_completion $(ARTIFACTDIR)/_zrepl.zsh_completion $(ARTIFACTDIR)/go_env.txt docs docs-clean
.PHONY: noarch $(ARTIFACTDIR)/bash_completion $(ARTIFACTDIR)/go_env.txt docs docs-clean
$(ARTIFACTDIR):
@@ -168,16 +168,12 @@ $(ARTIFACTDIR):
$(ARTIFACTDIR)/docs: $(ARTIFACTDIR)
mkdir -p "$@"
noarch: $(ARTIFACTDIR)/bash_completion $(ARTIFACTDIR)/_zrepl.zsh_completion $(ARTIFACTDIR)/go_env.txt docs
noarch: $(ARTIFACTDIR)/bash_completion $(ARTIFACTDIR)/go_env.txt docs
# pass
$(ARTIFACTDIR)/bash_completion:
$(MAKE) zrepl-bin GOOS=$(GOHOSTOS) GOARCH=$(GOHOSTARCH)
artifacts/zrepl-$(GOHOSTOS)-$(GOHOSTARCH) gencompletion bash "$@"
$(ARTIFACTDIR)/_zrepl.zsh_completion:
$(MAKE) zrepl-bin GOOS=$(GOHOSTOS) GOARCH=$(GOHOSTARCH)
artifacts/zrepl-$(GOHOSTOS)-$(GOHOSTARCH) gencompletion zsh "$@"
artifacts/zrepl-$(GOHOSTOS)-$(GOHOSTARCH) bashcomp "$@"
$(ARTIFACTDIR)/go_env.txt:
$(GO_ENV_VARS) $(GO) env > $@
+19 -45
View File
@@ -19,55 +19,29 @@ var rootCmd = &cobra.Command{
Short: "One-stop ZFS replication solution",
}
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",
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)
}
},
Hidden: true,
}
func init() {
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)
rootCmd.PersistentFlags().StringVar(&rootArgs.configPath, "config", "", "config file path")
rootCmd.AddCommand(bashcompCmd)
}
type Subcommand struct {
@@ -12,28 +12,28 @@ import (
"github.com/zrepl/zrepl/zfs"
)
var zabsCreateStepHoldFlags struct {
var holdsCreateStepHoldFlags struct {
target string
jobid JobIDFlag
}
var zabsCmdCreateStepHold = &cli.Subcommand{
var holdsCmdCreateStepHold = &cli.Subcommand{
Use: "step",
Run: doZabsCreateStep,
Run: doHoldsCreateStep,
NoRequireConfig: true,
Short: `create a step hold or bookmark`,
SetupFlags: func(f *pflag.FlagSet) {
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")
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")
},
}
func doZabsCreateStep(sc *cli.Subcommand, args []string) error {
func doHoldsCreateStep(sc *cli.Subcommand, args []string) error {
if len(args) > 0 {
return errors.New("subcommand takes no arguments")
}
f := &zabsCreateStepHoldFlags
f := &holdsCreateStepHoldFlags
fs, _, _, err := zfs.DecomposeVersionString(f.target)
if err != nil {
+10 -10
View File
@@ -14,15 +14,15 @@ import (
)
var (
ZFSAbstractionsCmd = &cli.Subcommand{
Use: "zfs-abstraction",
Short: "manage abstractions that zrepl builds on top of ZFS",
HoldsCmd = &cli.Subcommand{
Use: "holds",
Short: "manage holds & step bookmarks",
SetupSubcommands: func() []*cli.Subcommand {
return []*cli.Subcommand{
zabsCmdList,
zabsCmdReleaseAll,
zabsCmdReleaseStale,
zabsCmdCreate,
holdsCmdList,
holdsCmdReleaseAll,
holdsCmdReleaseStale,
holdsCmdCreate,
}
},
}
@@ -30,7 +30,7 @@ var (
// a common set of CLI flags that map to the fields of an
// endpoint.ListZFSHoldsAndBookmarksQuery
type zabsFilterFlags struct {
type holdsFilterFlags struct {
Filesystems FilesystemsFilterFlag
Job JobIDFlag
Types AbstractionTypesFlag
@@ -38,7 +38,7 @@ type zabsFilterFlags struct {
}
// produce a query from the CLI flags
func (f zabsFilterFlags) Query() (endpoint.ListZFSHoldsAndBookmarksQuery, error) {
func (f holdsFilterFlags) Query() (endpoint.ListZFSHoldsAndBookmarksQuery, error) {
q := endpoint.ListZFSHoldsAndBookmarksQuery{
FS: f.Filesystems.FlagValue(),
What: f.Types.FlagValue(),
@@ -48,7 +48,7 @@ func (f zabsFilterFlags) Query() (endpoint.ListZFSHoldsAndBookmarksQuery, error)
return q, q.Validate()
}
func (f *zabsFilterFlags) registerZabsFilterFlags(s *pflag.FlagSet, verb string) {
func (f *holdsFilterFlags) registerHoldsFilterFlags(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))
+14
View File
@@ -0,0 +1,14 @@
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,
}
},
}
@@ -16,23 +16,23 @@ import (
"github.com/zrepl/zrepl/util/chainlock"
)
var zabsListFlags struct {
Filter zabsFilterFlags
var holdsListFlags struct {
Filter holdsFilterFlags
Json bool
}
var zabsCmdList = &cli.Subcommand{
var holdsCmdList = &cli.Subcommand{
Use: "list",
Short: `list zrepl ZFS abstractions`,
Run: doZabsList,
Run: doHoldsList,
NoRequireConfig: true,
Short: "list holds and bookmarks",
SetupFlags: func(f *pflag.FlagSet) {
zabsListFlags.Filter.registerZabsFilterFlags(f, "list")
f.BoolVar(&zabsListFlags.Json, "json", false, "emit JSON")
holdsListFlags.Filter.registerHoldsFilterFlags(f, "list")
f.BoolVar(&holdsListFlags.Json, "json", false, "emit JSON")
},
}
func doZabsList(sc *cli.Subcommand, args []string) error {
func doHoldsList(sc *cli.Subcommand, args []string) error {
var err error
ctx := context.Background()
@@ -40,7 +40,7 @@ func doZabsList(sc *cli.Subcommand, args []string) error {
return errors.New("this subcommand takes no positional arguments")
}
q, err := zabsListFlags.Filter.Query()
q, err := holdsListFlags.Filter.Query()
if err != nil {
return errors.Wrap(err, "invalid filter specification on command line")
}
@@ -62,7 +62,7 @@ func doZabsList(sc *cli.Subcommand, args []string) error {
for a := range abstractions {
func() {
defer line.Lock().Unlock()
if zabsListFlags.Json {
if holdsListFlags.Json {
enc.SetIndent("", " ")
if err := enc.Encode(abstractions); err != nil {
panic(err)
@@ -15,35 +15,35 @@ import (
)
// shared between release-all and release-step
var zabsReleaseFlags struct {
Filter zabsFilterFlags
var holdsReleaseFlags struct {
Filter holdsFilterFlags
Json bool
DryRun bool
}
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")
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")
}
var zabsCmdReleaseAll = &cli.Subcommand{
var holdsCmdReleaseAll = &cli.Subcommand{
Use: "release-all",
Run: doZabsReleaseAll,
Run: doHoldsReleaseAll,
NoRequireConfig: true,
Short: `(DANGEROUS) release ALL zrepl ZFS abstractions (mostly useful for uninstalling zrepl completely or for "de-zrepl-ing" a filesystem)`,
SetupFlags: registerZabsReleaseFlags,
Short: `(DANGEROUS) release all zrepl-managed holds and bookmarks, mostly useful for uninstalling zrepl`,
SetupFlags: registerHoldsReleaseFlags,
}
var zabsCmdReleaseStale = &cli.Subcommand{
var holdsCmdReleaseStale = &cli.Subcommand{
Use: "release-stale",
Run: doZabsReleaseStale,
Run: doHoldsReleaseStale,
NoRequireConfig: true,
Short: `release stale zrepl ZFS abstractions (useful if zrepl has a bug and does not do it by itself)`,
SetupFlags: registerZabsReleaseFlags,
Short: `release stale zrepl-managed holds and boomkarks (useful if zrepl has a bug and doesn't do it by itself)`,
SetupFlags: registerHoldsReleaseFlags,
}
func doZabsReleaseAll(sc *cli.Subcommand, args []string) error {
func doHoldsReleaseAll(sc *cli.Subcommand, args []string) error {
var err error
ctx := context.Background()
@@ -51,7 +51,7 @@ func doZabsReleaseAll(sc *cli.Subcommand, args []string) error {
return errors.New("this subcommand takes no positional arguments")
}
q, err := zabsReleaseFlags.Filter.Query()
q, err := holdsReleaseFlags.Filter.Query()
if err != nil {
return errors.Wrap(err, "invalid filter specification on command line")
}
@@ -65,10 +65,10 @@ func doZabsReleaseAll(sc *cli.Subcommand, args []string) error {
// proceed anyways with rest of abstractions
}
return doZabsRelease_Common(ctx, abstractions)
return doHoldsRelease_Common(ctx, abstractions)
}
func doZabsReleaseStale(sc *cli.Subcommand, args []string) error {
func doHoldsReleaseStale(sc *cli.Subcommand, args []string) error {
var err error
ctx := context.Background()
@@ -77,7 +77,7 @@ func doZabsReleaseStale(sc *cli.Subcommand, args []string) error {
return errors.New("this subcommand takes no positional arguments")
}
q, err := zabsReleaseFlags.Filter.Query()
q, err := holdsReleaseFlags.Filter.Query()
if err != nil {
return errors.Wrap(err, "invalid filter specification on command line")
}
@@ -87,13 +87,13 @@ func doZabsReleaseStale(sc *cli.Subcommand, args []string) error {
return err // context clear by invocation of command
}
return doZabsRelease_Common(ctx, stalenessInfo.Stale)
return doHoldsRelease_Common(ctx, stalenessInfo.Stale)
}
func doZabsRelease_Common(ctx context.Context, destroy []endpoint.Abstraction) error {
func doHoldsRelease_Common(ctx context.Context, destroy []endpoint.Abstraction) error {
if zabsReleaseFlags.DryRun {
if zabsReleaseFlags.Json {
if holdsReleaseFlags.DryRun {
if holdsReleaseFlags.Json {
m, err := json.MarshalIndent(destroy, "", " ")
if err != nil {
panic(err)
@@ -121,7 +121,7 @@ func doZabsRelease_Common(ctx context.Context, destroy []endpoint.Abstraction) e
for res := range outcome {
hadErr = hadErr || res.DestroyErr != nil
if zabsReleaseFlags.Json {
if holdsReleaseFlags.Json {
err := enc.Encode(res)
if err != nil {
colorErr.Fprintf(os.Stderr, "cannot marshal there were errors in destroying the abstractions")
-14
View File
@@ -1,14 +0,0 @@
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,
}
},
}
-2
View File
@@ -62,9 +62,7 @@ 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
-----
+2 -2
View File
@@ -38,8 +38,8 @@ CLI Overview
* - ``zrepl migrate``
- | perform on-disk state / ZFS property migrations
| (see :ref:`changelog <changelog>` for details)
* - ``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>` )
* - ``zrepl holds``
- list and remove holds and step bookmarks created by zrepl (see :ref:`overview <replication-cursor-and-last-received-hold>` )
.. _usage-zrepl-daemon:
+1 -1
View File
@@ -806,7 +806,7 @@ func (s *Receiver) Receive(ctx context.Context, req *pdu.ReceiveReq, receive zfs
if err := zfs.ZFSRecv(ctx, lp.ToString(), to, receive, recvOpts); err != nil {
getLogger(ctx).
WithError(err).
WithField("opts", recvOpts).
WithField("opts", fmt.Sprintf("%#v", recvOpts)).
Error("zfs receive failed")
return nil, err
}
+1 -1
View File
@@ -17,7 +17,7 @@ func init() {
cli.AddSubcommand(client.PprofCmd)
cli.AddSubcommand(client.TestCmd)
cli.AddSubcommand(client.MigrateCmd)
cli.AddSubcommand(client.ZFSAbstractionsCmd)
cli.AddSubcommand(client.HoldsCmd)
}
func main() {
+34 -31
View File
@@ -355,7 +355,7 @@ func (a *attempt) do(ctx context.Context, prev *attempt) {
// invariant: prevs contains an entry for each unambiguous correspondence
stepQueue := newStepQueue()
defer stepQueue.Start(1)() // TODO parallel replication
defer stepQueue.Start(envconst.Int("ZREPL_REPLICATION_EXPERIMENTAL_REPLICATION_CONCURRENCY", 1))() // TODO parallel replication
var fssesDone sync.WaitGroup
for _, f := range a.fss {
fssesDone.Add(1)
@@ -370,54 +370,57 @@ func (a *attempt) do(ctx context.Context, prev *attempt) {
a.finishedAt = time.Now()
}
func (fs *fs) do(ctx context.Context, pq *stepQueue, prev *fs) {
func (f *fs) debug(format string, args ...interface{}) {
debugPrefix("fs=%s", f.fs.ReportInfo().Name)(format, args...)
}
defer fs.l.Lock().Unlock()
func (f *fs) do(ctx context.Context, pq *stepQueue, prev *fs) {
defer f.l.Lock().Unlock()
// get planned steps from replication logic
var psteps []Step
var errTime time.Time
var err error
fs.l.DropWhile(func() {
f.l.DropWhile(func() {
// TODO hacky
// choose target time that is earlier than any snapshot, so fs planning is always prioritized
targetDate := time.Unix(0, 0)
defer pq.WaitReady(fs, targetDate)()
psteps, err = fs.fs.PlanFS(ctx) // no shadow
errTime = time.Now() // no shadow
defer pq.WaitReady(f, targetDate)()
psteps, err = f.fs.PlanFS(ctx) // no shadow
errTime = time.Now() // no shadow
})
debug := debugPrefix("fs=%s", fs.fs.ReportInfo().Name)
fs.planning.done = true
f.planning.done = true
if err != nil {
fs.planning.err = newTimedError(err, errTime)
f.planning.err = newTimedError(err, errTime)
return
}
for _, pstep := range psteps {
step := &step{
l: fs.l,
l: f.l,
step: pstep,
}
fs.planned.steps = append(fs.planned.steps, step)
f.planned.steps = append(f.planned.steps, step)
}
debug("initial len(fs.planned.steps) = %d", len(fs.planned.steps))
f.debug("initial len(fs.planned.steps) = %d", len(f.planned.steps))
// for not-first attempts, only allow fs.planned.steps
// up to including the originally planned target snapshot
if prev != nil && prev.planning.done && prev.planning.err == nil {
prevUncompleted := prev.planned.steps[prev.planned.step:]
if len(prevUncompleted) == 0 {
debug("prevUncompleted is empty")
f.debug("prevUncompleted is empty")
return
}
if len(fs.planned.steps) == 0 {
debug("fs.planned.steps is empty")
if len(f.planned.steps) == 0 {
f.debug("fs.planned.steps is empty")
return
}
prevFailed := prevUncompleted[0]
curFirst := fs.planned.steps[0]
curFirst := f.planned.steps[0]
// we assume that PlanFS retries prevFailed (using curFirst)
if !prevFailed.step.TargetEquals(curFirst.step) {
debug("Targets don't match")
f.debug("Targets don't match")
// Two options:
// A: planning algorithm is broken
// B: manual user intervention inbetween
@@ -433,43 +436,43 @@ func (fs *fs) do(ctx context.Context, pq *stepQueue, prev *fs) {
}
msg := fmt.Sprintf("last attempt's uncompleted step %s does not correspond to this attempt's first planned step %s",
stepFmt(prevFailed), stepFmt(curFirst))
fs.planned.stepErr = newTimedError(errors.New(msg), time.Now())
f.planned.stepErr = newTimedError(errors.New(msg), time.Now())
return
}
// only allow until step targets diverge
min := len(prevUncompleted)
if min > len(fs.planned.steps) {
min = len(fs.planned.steps)
if min > len(f.planned.steps) {
min = len(f.planned.steps)
}
diverge := 0
for ; diverge < min; diverge++ {
debug("diverge compare iteration %d", diverge)
if !fs.planned.steps[diverge].step.TargetEquals(prevUncompleted[diverge].step) {
f.debug("diverge compare iteration %d", diverge)
if !f.planned.steps[diverge].step.TargetEquals(prevUncompleted[diverge].step) {
break
}
}
debug("diverge is %d", diverge)
fs.planned.steps = fs.planned.steps[0:diverge]
f.debug("diverge is %d", diverge)
f.planned.steps = f.planned.steps[0:diverge]
}
debug("post-prev-merge len(fs.planned.steps) = %d", len(fs.planned.steps))
f.debug("post-prev-merge len(fs.planned.steps) = %d", len(f.planned.steps))
for i, s := range fs.planned.steps {
for i, s := range f.planned.steps {
var (
err error
errTime time.Time
)
// lock must not be held while executing step in order for reporting to work
fs.l.DropWhile(func() {
f.l.DropWhile(func() {
targetDate := s.step.TargetDate()
defer pq.WaitReady(fs, targetDate)()
defer pq.WaitReady(f, targetDate)()
err = s.step.Step(ctx) // no shadow
errTime = time.Now() // no shadow
})
if err != nil {
fs.planned.stepErr = newTimedError(err, errTime)
f.planned.stepErr = newTimedError(err, errTime)
break
}
fs.planned.step = i + 1 // fs.planned.step must be == len(fs.planned.steps) if all went OK
f.planned.step = i + 1 // fs.planned.step must be == len(fs.planned.steps) if all went OK
}
}
+3 -16
View File
@@ -1,7 +1,6 @@
package zfscmd
import (
"os/exec"
"time"
)
@@ -29,22 +28,10 @@ func waitPreLogging(c *Cmd, now time.Time) {
}
func waitPostLogging(c *Cmd, err error, now time.Time) {
var total, system, user float64
total = c.Runtime().Seconds()
if ee, ok := err.(*exec.ExitError); ok {
system = ee.ProcessState.SystemTime().Seconds()
user = ee.ProcessState.UserTime().Seconds()
} else {
system = -1
user = -1
}
log := c.log().
WithField("total_time_s", total).
WithField("systemtime_s", system).
WithField("usertime_s", user)
WithField("total_time_s", c.Runtime().Seconds()).
WithField("systemtime_s", c.cmd.ProcessState.SystemTime().Seconds()).
WithField("usertime_s", c.cmd.ProcessState.UserTime().Seconds())
if err == nil {
log.Info("command exited without error")
-29
View File
@@ -1,9 +1,7 @@
package zfscmd
import (
"bufio"
"bytes"
"context"
"io"
"os/exec"
"testing"
@@ -60,30 +58,3 @@ func TestCmdStderrBehaviorStdoutPipe(t *testing.T) {
require.True(t, ok)
require.Empty(t, ee.Stderr) // !!!!! probably not what one would expect if we only redirect stdout
}
func TestCmdProcessState(t *testing.T) {
ctx, cancel := context.WithCancel(context.Background())
defer cancel()
cmd := exec.CommandContext(ctx, "bash", "-c", "echo running; sleep 3600")
stdout, err := cmd.StdoutPipe()
require.NoError(t, err)
err = cmd.Start()
require.NoError(t, err)
r := bufio.NewReader(stdout)
line, err := r.ReadString('\n')
require.NoError(t, err)
require.Equal(t, "running\n", line)
// we know it's running and sleeping
cancel()
err = cmd.Wait()
t.Logf("wait err %T\n%s", err, err)
require.Error(t, err)
ee, ok := err.(*exec.ExitError)
require.True(t, ok)
require.NotNil(t, ee.ProcessState)
require.Contains(t, ee.Error(), "killed")
}