Make prune and remote nuke fail when a delete fails (closes #290)

prune, and the prune step of snapshot create --prune, counted a blob
it could not delete and still exited 0. It now prints its summary or
--json document and then returns an error, which replaces the
"Failed to delete N blob(s)" line.

remote nuke only warned when a snapshot's metadata could not be
deleted, advised vaultik prune, which never removes metadata, and then
reported the store empty. It now stops with an error, does not report
the store empty, and says to run vaultik remote nuke --force again. A
successful nuke no longer advises prune for blobs it deletes itself.

The faultstore test double gains an OnDelete hook.

Judgement call: a snapshot nuke cannot remove from the local index
also fails the run, since its metadata is then left on the store.

Model: opus-5-5

internal/storage/faultstore/faultstore.go

@@ -1,15 +1,16 @@
 // Package faultstore provides a storage.Storer wrapper that injects
 // faults on demand, so tests can reproduce the failure modes a real
 // backend exhibits: an upload that fails partway, a backend that reports
-// success while storing nothing, and reads that return corrupt or
-// truncated bytes. It is the seam called for by the fault-injection
-// tests (sneak/vaultik issue 72) and is meant to be reused by future
-// tests rather than re-implemented per case.
+// success while storing nothing, reads that return corrupt or truncated
+// bytes, and a delete the backend refuses. It is the seam called for by
+// the fault-injection tests (sneak/vaultik issue 72) and is meant to be
+// reused by future tests rather than re-implemented per case.
 //
-// The wrapper delegates every method to the inner Storer. Two hooks
-// change that: OnPut decides the fate of each write, and OnGet decides
-// how each read's bytes are returned. Both are keyed by the object key,
-// so a test can fault only blobs, only metadata, or a single object.
+// The wrapper delegates every method to the inner Storer. Three hooks
+// change that: OnPut decides the fate of each write, OnGet decides how
+// each read's bytes are returned, and OnDelete decides whether each
+// delete fails. All are keyed by the object key, so a test can fault
+// only blobs, only metadata, or a single object.
 package faultstore
 
 import (
@@ -25,6 +26,10 @@ import (
 // ErrInjectedUpload is returned by a Put the OnPut hook chose to fail.
 var ErrInjectedUpload = errors.New("faultstore: injected upload failure")
 
+// ErrInjectedDelete is returned by a Delete the OnDelete hook chose to
+// fail.
+var ErrInjectedDelete = errors.New("faultstore: injected delete failure")
+
 // PutAction is the disposition OnPut assigns to a write.
 type PutAction int
 
@@ -66,6 +71,11 @@ type Storer struct {
 	// OnGet, when set, is consulted for every Get with the object key
 	// and damages the returned bytes accordingly.
 	OnGet func(key string) GetFault
+
+	// OnDelete, when set, is consulted before every Delete with the
+	// object key. Returning true fails the delete and leaves the object
+	// in place.
+	OnDelete func(key string) bool
 }
 
 // New wraps inner. inner must be non-nil.
@@ -154,8 +164,12 @@ func (f *Storer) Stat(ctx context.Context, key string) (*storage.ObjectInfo, err
 	return f.inner.Stat(ctx, key)
 }
 
-// Delete delegates unchanged.
+// Delete removes the object unless OnDelete fails the delete.
 func (f *Storer) Delete(ctx context.Context, key string) error {
+	if f.OnDelete != nil && f.OnDelete(key) {
+		return fmt.Errorf("%w for %q", ErrInjectedDelete, key)
+	}
+
 	return f.inner.Delete(ctx, key)
 }
 

internal/vaultik/nuke_remote_test.go

@@ -5,12 +5,14 @@ import (
 	"io/fs"
 	"os"
 	"path/filepath"
+	"strings"
 	"testing"
 
 	"github.com/stretchr/testify/assert"
 	"github.com/stretchr/testify/require"
 	"sneak.berlin/go/vaultik/internal/log"
 	"sneak.berlin/go/vaultik/internal/snapshot"
+	"sneak.berlin/go/vaultik/internal/storage/faultstore"
 )
 
 // TestNukeRemoteLeavesNoFiles checks that remote nuke leaves no file
@@ -23,7 +25,7 @@ func TestNukeRemoteLeavesNoFiles(t *testing.T) {
 
 	ctx := context.Background()
 	storeDir := filepath.Join(t.TempDir(), "store")
-	v, repos, _ := backUpToFileDestination(ctx, t, storeDir)
+	v, repos, out := backUpToFileDestination(ctx, t, storeDir)
 
 	snapshots, err := repos.Snapshots.ListRecent(ctx, listRecentTestLimit)
 	require.NoError(t, err)
@@ -61,4 +63,62 @@ func TestNukeRemoteLeavesNoFiles(t *testing.T) {
 		})
 	require.NoError(t, err)
 	assert.Empty(t, files)
+
+	// Nuke deletes the blobs itself, so advice to run prune is wrong.
+	assert.Contains(t, out.String(), "Backup destination store is now empty.")
+	assert.NotContains(t, out.String(), "vaultik prune")
+}
+
+// TestNukeRemoteFailsWhenADeleteFails checks that remote nuke returns an
+// error when a snapshot's metadata or a blob cannot be deleted, does not
+// report the store empty, and advises running nuke again: prune deletes
+// no metadata.
+func TestNukeRemoteFailsWhenADeleteFails(t *testing.T) {
+	log.Initialize(log.Config{})
+	t.Parallel()
+
+	for _, refusedPrefix := range []string{"metadata/", "blobs/"} {
+		t.Run(refusedPrefix, func(t *testing.T) {
+			t.Parallel()
+
+			ctx := context.Background()
+			storeDir := filepath.Join(t.TempDir(), "store")
+			v, _, out := backUpToFileDestination(ctx, t, storeDir)
+
+			store := faultstore.New(v.Storage)
+			store.OnDelete = func(key string) bool {
+				return strings.HasPrefix(key, refusedPrefix)
+			}
+			v.Storage = store
+
+			require.Error(t, v.NukeRemote(true))
+
+			assert.NotContains(t, out.String(), "now empty")
+			assert.Contains(t, out.String(), "vaultik remote nuke --force")
+			assert.NotContains(t, out.String(), "vaultik prune")
+		})
+	}
+}
+
+// TestNukeRemoteFailsWhenALocalSnapshotCannotBeRemoved checks that a
+// snapshot remote nuke cannot remove from the local index fails the
+// run: nuke then leaves that snapshot's metadata on the destination
+// store.
+func TestNukeRemoteFailsWhenALocalSnapshotCannotBeRemoved(t *testing.T) {
+	log.Initialize(log.Config{})
+	t.Parallel()
+
+	ctx := context.Background()
+	storeDir := filepath.Join(t.TempDir(), "store")
+	v, _, out := backUpToFileDestination(ctx, t, storeDir)
+
+	_, err := v.DB.Conn().ExecContext(ctx, `
+		CREATE TRIGGER refuse_snapshot_delete BEFORE DELETE ON snapshots
+		BEGIN SELECT RAISE(ABORT, 'simulated index error'); END`)
+	require.NoError(t, err)
+
+	require.Error(t, v.NukeRemote(true))
+
+	assert.NotContains(t, out.String(), "now empty")
+	assert.Contains(t, out.String(), "vaultik remote nuke --force")
 }

internal/vaultik/prune.go

@@ -19,6 +19,14 @@ type PruneOptions struct {
 var errNukeRequiresForce = errors.New(
 	"nuke requires --force (this deletes ALL remote snapshots and blobs)")
 
+// errBlobDeleteFailed reports unreferenced blobs prune could not delete.
+var errBlobDeleteFailed = errors.New("blob(s) could not be deleted")
+
+// nukeCommandHint is the command suggested when a remote nuke could not
+// delete everything: running it again deletes what is left, which
+// `vaultik prune` does not do for snapshot metadata.
+const nukeCommandHint = "vaultik remote nuke --force"
+
 // metadataDirName is the top-level remote directory holding snapshot
 // metadata.
 const metadataDirName = "metadata"
@@ -35,6 +43,22 @@ func (v *Vaultik) NukeRemote(force bool) error {
 		return errNukeRequiresForce
 	}
 
+	err := v.emptyDestinationStore()
+	if err != nil {
+		v.UI.Warningf("Run '%s' again to delete what is left on the "+
+			"backup destination store.", nukeCommandHint)
+
+		return err
+	}
+
+	v.UI.Completef("Backup destination store is now empty.")
+
+	return nil
+}
+
+// emptyDestinationStore does the deleting for NukeRemote and stops at
+// the first step that fails.
+func (v *Vaultik) emptyDestinationStore() error {
 	v.UI.Beginf("Removing all snapshot metadata from backup destination store.")
 
 	_, err := v.RemoveAllSnapshots(&RemoveOptions{Force: true})
@@ -58,8 +82,6 @@ func (v *Vaultik) NukeRemote(force bool) error {
 		}
 	}
 
-	v.UI.Completef("Backup destination store is now empty.")
-
 	return nil
 }
 
@@ -162,14 +184,17 @@ func (v *Vaultik) PruneBlobs(opts *PruneOptions) error {
 	v.deleteUnreferencedBlobs(unreferencedBlobs, allBlobs, result)
 
 	if opts.JSON {
-		return v.outputPruneBlobsJSON(result)
+		err = v.outputPruneBlobsJSON(result)
+		if err != nil {
+			return err
+		}
+	} else {
+		v.stdoutf("\nDeleted %d blob(s) totaling %s\n",
+			result.BlobsDeleted, ubytes(result.BytesFreed))
 	}
 
-	v.stdoutf("\nDeleted %d blob(s) totaling %s\n",
-		result.BlobsDeleted, ubytes(result.BytesFreed))
-
 	if result.BlobsFailed > 0 {
-		v.stdoutf("Failed to delete %d blob(s)\n", result.BlobsFailed)
+		return fmt.Errorf("%d %w", result.BlobsFailed, errBlobDeleteFailed)
 	}
 
 	return nil

internal/vaultik/prune_delete_failure_test.go

@@ -0,0 +1,98 @@
+package vaultik_test
+
+import (
+	"context"
+	"encoding/json"
+	"os"
+	"path/filepath"
+	"strings"
+	"testing"
+
+	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
+	"sneak.berlin/go/vaultik/internal/log"
+	"sneak.berlin/go/vaultik/internal/storage/faultstore"
+	"sneak.berlin/go/vaultik/internal/vaultik"
+)
+
+// refuseBlobDeletes is an OnDelete hook that fails every blob delete.
+func refuseBlobDeletes(key string) bool {
+	return strings.HasPrefix(key, "blobs/")
+}
+
+// TestPruneBlobs_FailedDeleteReturnsError checks that prune returns an
+// error, so the command exits non-zero, when an unreferenced blob cannot
+// be deleted.
+func TestPruneBlobs_FailedDeleteReturnsError(t *testing.T) {
+	log.Initialize(log.Config{})
+	t.Parallel()
+
+	env := newListEnv(t)
+
+	// The store holds no manifest, so nothing references this blob.
+	addBlob(t, env.store.testStorer, testBlobHashA)
+
+	store := faultstore.New(env.store)
+	store.OnDelete = refuseBlobDeletes
+	env.v.Storage = store
+
+	err := env.v.PruneBlobs(&vaultik.PruneOptions{Force: true})
+
+	require.ErrorContains(t, err, "1 blob(s) could not be deleted")
+}
+
+// TestPruneBlobs_JSONFailedDeleteReturnsError checks that prune --json
+// prints its document and then returns an error when an unreferenced
+// blob cannot be deleted.
+func TestPruneBlobs_JSONFailedDeleteReturnsError(t *testing.T) {
+	log.Initialize(log.Config{})
+	t.Parallel()
+
+	env := newListEnv(t)
+	addBlob(t, env.store.testStorer, testBlobHashA)
+
+	store := faultstore.New(env.store)
+	store.OnDelete = refuseBlobDeletes
+	env.v.Storage = store
+
+	err := env.v.PruneBlobs(&vaultik.PruneOptions{JSON: true})
+
+	require.Error(t, err, "prune must fail when a blob cannot be deleted")
+
+	var result vaultik.PruneBlobsResult
+
+	require.NoError(t, json.Unmarshal(env.stdout.Bytes(), &result),
+		"stdout must hold the JSON document, got:\n%s", env.stdout.String())
+	assert.Equal(t, 1, result.BlobsFailed)
+	assert.Equal(t, 0, result.BlobsDeleted)
+}
+
+// TestCreateSnapshotPrune_FailedBlobDeleteReturnsError checks that the
+// prune step of snapshot create --prune fails the run when an
+// unreferenced blob cannot be deleted.
+func TestCreateSnapshotPrune_FailedBlobDeleteReturnsError(t *testing.T) {
+	log.Initialize(log.Config{})
+	t.Parallel()
+
+	ctx := context.Background()
+	storeDir := filepath.Join(t.TempDir(), "store")
+	v, _, _ := backUpToFileDestination(ctx, t, storeDir)
+
+	// No snapshot references this blob, so the prune step deletes it.
+	orphan := filepath.Join(storeDir, "blobs", testBlobHashA[:2],
+		testBlobHashA[2:4], testBlobHashA)
+	require.NoError(t, os.MkdirAll(filepath.Dir(orphan), 0o750))
+	require.NoError(t, os.WriteFile(orphan, []byte("orphan"), 0o600))
+
+	store := faultstore.New(v.Storage)
+	store.OnDelete = refuseBlobDeletes
+	v.Storage = store
+
+	err := v.CreateSnapshot(&vaultik.SnapshotCreateOptions{
+		Cron:  true,
+		Prune: true,
+	})
+
+	require.Error(t, err, "the prune step must fail the run")
+	assert.FileExists(t, orphan)
+}

internal/vaultik/remove_snapshot_test.go

@@ -361,7 +361,8 @@ func TestRemoveAllSnapshots_WithForce(t *testing.T) {
 	out := tv.Stdout.String()
 	assert.Contains(t, out, "Removed 2 snapshot(s)")
 	assert.Contains(t, out, "Removed snapshot metadata from remote storage")
-	assert.Contains(t, out, "vaultik prune")
+	// Remote nuke, the only caller, deletes the blobs next itself.
+	assert.NotContains(t, out, "vaultik prune")
 }
 
 func TestRemoveAllSnapshots_DryRun(t *testing.T) {

internal/vaultik/snapshot.go

@@ -23,6 +23,7 @@ var (
 	errNoSnapshotsInConfig  = errors.New("no snapshots configured")
 	errSnapshotVerifyFailed = errors.New("verification failed")
 	errRemoveAllNeedsForce  = errors.New("--all requires --force")
+	errSnapshotRemoveFailed = errors.New("snapshot(s) could not be removed")
 	errInvalidTableName     = errors.New("invalid table name")
 	errNoAgeRecipients      = errors.New(
 		"creating a snapshot needs at least one public key in " +
@@ -1464,8 +1465,8 @@ func (v *Vaultik) handleRemoveAllDryRun(
 
 // executeRemoveAll deletes every local snapshot and, unless LocalOnly
 // is set, every corresponding remote metadata directory plus any
-// orphan remote keys. Blobs are not touched; the printed prune-command
-// hint is the next step.
+// orphan remote keys. Blobs are not touched. It returns an error when
+// any snapshot could not be removed.
 func (v *Vaultik) executeRemoveAll(
 	localSnaps, orphanRemoteKeys []string, opts *RemoveOptions,
 ) (*RemoveResult, error) {
@@ -1479,18 +1480,11 @@ func (v *Vaultik) executeRemoveAll(
 
 	result := &RemoveResult{}
 
-	remoteErrors := v.removeAllLocalSnapshots(localSnaps, opts, result)
+	failures := v.removeAllLocalSnapshots(localSnaps, opts, result)
 
 	if !opts.LocalOnly {
-		remoteErrors += v.removeAllOrphanRemotes(orphanRemoteKeys, result)
-
-		if remoteErrors == 0 {
-			result.RemoteRemoved = true
-		} else if v.UI != nil {
-			v.UI.Warningf("Some remote metadata deletions failed. Run '%s' "+
-				"once the remote is healthy to clean up unreferenced blobs.",
-				pruneCommandHint)
-		}
+		failures += v.removeAllOrphanRemotes(orphanRemoteKeys, result)
+		result.RemoteRemoved = failures == 0
 	}
 
 	if v.SnapshotManager != nil {
@@ -1500,6 +1494,10 @@ func (v *Vaultik) executeRemoveAll(
 		}
 	}
 
+	if failures > 0 {
+		return result, fmt.Errorf("%d %w", failures, errSnapshotRemoveFailed)
+	}
+
 	if opts.JSON {
 		return result, v.outputRemoveJSON(result)
 	}
@@ -1508,9 +1506,6 @@ func (v *Vaultik) executeRemoveAll(
 
 	if !opts.LocalOnly && result.RemoteRemoved {
 		v.printlnStdout("Removed snapshot metadata from remote storage")
-		v.stdoutf("\nNote: Removed snapshots' blobs remain on the remote. "+
-			"Run '%s' to delete any blobs no longer referenced by any "+
-			"remaining remote snapshot.\n", pruneCommandHint)
 	}
 
 	return result, nil
@@ -1553,12 +1548,13 @@ func (v *Vaultik) deleteIncompleteSnapshot(snapshotIDStr string, id any) bool {
 }
 
 // removeAllLocalSnapshots removes each known local snapshot (and, unless
-// --local-only, its remote metadata), returning the number of remote
-// deletion failures.
+// --local-only, its remote metadata), returning the number of snapshots
+// it could not remove. A snapshot left in the local index keeps its
+// remote metadata too, since it is not an orphan remote key.
 func (v *Vaultik) removeAllLocalSnapshots(
 	localSnaps []string, opts *RemoveOptions, result *RemoveResult,
 ) int {
-	remoteErrors := 0
+	failures := 0
 
 	for _, snapshotID := range localSnaps {
 		log.Info("Removing snapshot", "snapshot_id", snapshotID)
@@ -1568,6 +1564,8 @@ func (v *Vaultik) removeAllLocalSnapshots(
 			log.Error("Failed to remove from local database",
 				"snapshot_id", snapshotID, "error", err)
 
+			failures++
+
 			continue
 		}
 
@@ -1578,14 +1576,14 @@ func (v *Vaultik) removeAllLocalSnapshots(
 				log.Warn("Failed to remove snapshot metadata from remote",
 					"snapshot_id", snapshotID, "error", err)
 
-				remoteErrors++
+				failures++
 			}
 		}
 
 		result.SnapshotsRemoved = append(result.SnapshotsRemoved, snapshotID)
 	}
 
-	return remoteErrors
+	return failures
 }
 
 // removeAllOrphanRemotes deletes remote snapshot metadata that has no