[#316] endpoint / replication protocol: more robust step-holds and replication cursor management

- drop HintMostRecentCommonAncestor rpc call
    - it is wrong to put faith into the active side of the replication to always make that call
      (we might not trust it, ref pull setup)
- clean up step holds + step bookmarks + replication cursor bookmarks on
  send RPC instead
    - this makes it symmetric with Receive RPC
- use a cache (endpoint.sendAbstractionsCache) to avoid the cost of
  listing the on-disk endpoint abstractions state on every step

The "create" methods for endpoint abstractions (CreateReplicationCursor, HoldStep) are now fully
idempotent and return an Abstraction.

Notes about endpoint.sendAbstractionsCache:
- fills lazily from disk state on first `Get` operation
- fill from disk is generally only attempted once
    - unless the `ListAbstractions` fails, in which case the fill from
      disk is retried on next `Get` (the current `Get` will observe a
      subset of the actual on-disk abstractions)
    - the `Invalidate` method is called
- it is a global (zrepl process-wide) cache

fixes #316

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/replicationCursor.go

@@ -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")
@@ -32,9 +32,31 @@ func ReplicationCursor(ctx *platformtest.Context) {
 	}
 
 	fs := ds.ToString()
-	snap := fsversion(ctx, fs, "@1 with space")
 
-	destroyed, err := endpoint.MoveReplicationCursor(ctx, fs, &snap, jobid)
+	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)
+	}
+
+	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, cursorOfCursor.GetFilesystemVersion())
+
+	destroyed, err := endpoint.DestroyObsoleteReplicationCursors(ctx, fs, &snap, jobid)
 	if err != nil {
 		panic(err)
 	}
@@ -61,7 +83,11 @@ func ReplicationCursor(ctx *platformtest.Context) {
 	require.NoError(ctx, err)
 
 	snap2 := fsversion(ctx, fs, "@2 with space")
-	destroyed, err = endpoint.MoveReplicationCursor(ctx, fs, &snap2, jobid)
+
+	_, err = endpoint.CreateReplicationCursor(ctx, fs, snap, jobid)
+	assert.NoError(ctx, err)
+	destroyed, err = endpoint.DestroyObsoleteReplicationCursors(ctx, fs, &snap2, jobid)
+
 	require.NoError(ctx, err)
 	require.Equal(ctx, 1, len(destroyed))
 	require.Equal(ctx, endpoint.AbstractionReplicationCursorBookmarkV2, destroyed[0].GetType())