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. It
checks that the local index belongs to the configured destination
before deleting anything, and that error carries no such advice. A
successful nuke no longer advises prune.

Model: opus-5-5

README.md

@@ -349,9 +349,11 @@ the backup destination store. Blobs are NOT touched — deleting blobs
 requires reading every remaining remote manifest (the destination store
 may hold snapshots this host doesn't know about), which is what
 `vaultik prune` does. On success the command prints the exact `vaultik
-prune` invocation to run as a follow-up. Local row cleanup (files,
+prune` invocation to run as a follow-up. An ID that names no snapshot in
-chunks, blobs the snapshot was the last referrer for) runs
+the local index or on the destination store is an error; with
-automatically. If the destination store is unreachable, the local-DB
+`--local-only`, the snapshot must be in the local index. Local row
+cleanup (files, chunks, blobs the snapshot was the last referrer for)
+runs automatically. If the destination store is unreachable, the local-DB
 removal still completes and a warning is emitted; run `vaultik snapshot
 remove <snapshot-id>` again once the store is reachable to remove the
 snapshot's metadata from it (`vaultik prune` does not). To wipe everything
@@ -382,10 +384,12 @@ manifest on the destination store to compute the still-referenced blob
 set and delete any blob not in that set. Step (3) reads all remote
 manifests — network cost scales with the number of snapshots. `snapshot
 create --prune` runs the same cleanup automatically; this is the
-manual entry point for the same work.
+manual entry point for the same work. If a blob cannot be deleted, the
+command (and `snapshot create --prune`) exits non-zero.
 * `--force`: Skip confirmation prompt
 * `--json`: Output stats as JSON. Also skips the confirmation prompt, as
-  `--force` does.
+  `--force` does. A blob that cannot be deleted is counted in
+  `blobs_failed`, and the document is printed before the non-zero exit.
 
 **`info`**: Display system configuration, storage settings, encryption
 recipients, and local database statistics.
@@ -405,14 +409,16 @@ snapshot's blob count and blob size unknown, `null` in `--json`.
 **`remote nuke`**: Delete every snapshot's metadata and every blob from the
 backup destination store, leaving the bucket prefix empty. Destructive and
 irreversible. This is the single supported way to wipe the entire
-destination store.
+destination store. If anything cannot be deleted, the command exits
+non-zero; running it again deletes what is left.
 * `--force`: Required to confirm destruction.
 
 **`database delete`**: Delete the local SQLite state database file
 entirely. Remote storage is unaffected; the next backup will do a full
 scan and re-deduplicate against existing remote blobs, and the local
 index will re-bind to the currently configured storage destination.
-Use this after changing `storage_url` to a different destination.
+Use this after changing `storage_url`, or `s3.endpoint`, `s3.bucket` or
+`s3.prefix`, to a different destination.
 * `--force`: Skip confirmation prompt
 
 ---

TODO.md

@@ -22,6 +22,70 @@ the tag exists and is exercised; what is left is merging `next` to
 
 # Completed Steps
 
+- 2026-10-13: Made `prune` and `remote nuke` exit non-zero when a delete
+  fails ([issue #290](https://git.eeqj.de/sneak/vaultik/issues/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 fails. `remote nuke` warned about a
+  snapshot's metadata it could not delete, said to run `vaultik prune`,
+  which never removes metadata, and then reported the store empty. It
+  now fails without that report and says to run `vaultik remote nuke
+  --force` again, and a successful nuke no longer advises `prune` for
+  blobs it deletes itself.
+
+- 2026-10-13: Made `snapshot remove` fail for a snapshot that does not
+  exist ([issue #293](https://git.eeqj.de/sneak/vaultik/issues/293)).
+  The command never checked the ID: it reported the snapshot removed
+  from the local database and from the destination store and exited 0,
+  under `--json`, `--dry-run` and `--local-only` too. An ID that names
+  no snapshot in the local index or on the destination store is now an
+  error; with `--local-only`, the snapshot must be in the local index.
+  The command now reports a removal from either side only when one
+  happened, and `remote_removed` is true only when metadata was deleted.
+
+- 2026-10-12: Bound the local index to the destination an `s3.*` config
+  names ([issue #298](https://git.eeqj.de/sneak/vaultik/issues/298)).
+  Only `storage_url` was checked, so after `s3.bucket` or `s3.prefix`
+  changed, the next backup treated every chunk as already stored,
+  uploaded nothing, and completed a snapshot that could not be restored.
+  Without `storage_url`, the local index is now bound to `s3.endpoint`,
+  `s3.bucket` and `s3.prefix`, and a change is refused as a changed
+  `storage_url` is.
+
+- 2026-10-12: Rejected a retention duration longer than about 292 years
+  ([issue #289](https://git.eeqj.de/sneak/vaultik/issues/289)). The
+  number given to `--older-than` or `--keep-newer-than` was multiplied
+  by the length of its unit with no limit, so from `293y`, `15251w`,
+  `3559mo` or `106752d`, or a sum of parts such as `200y200y`, the
+  result wrapped to a negative duration. The cutoff then lay in the
+  future, every snapshot of the name was deleted, the one just made
+  included, and the command exited 0. Such a duration is now rejected
+  as invalid and nothing is deleted.
+
+- 2026-10-08: Made a backup cancelled under `--skip-errors` stop at once
+  ([issue #286](https://git.eeqj.de/sneak/vaultik/issues/286)). After
+  Ctrl-C or SIGTERM, phase 2 treated the cancellation error from each
+  remaining file like an unreadable file: it opened the file, printed an
+  error line for it, counted it as failed and went on to the next one.
+  The run now stops at the first file after the cancellation, and no
+  file is counted as failed because of it.
+
+- 2026-10-08: Made `remote nuke` delete the `.partial` files that
+  uploads cut off part-way leave on the destination store
+  ([issue #281](https://git.eeqj.de/sneak/vaultik/issues/281)). The
+  file and rclone listings skip such a file, so the command left it in
+  place and still reported the store empty. It now removes them under
+  `metadata/` and `blobs/` as its last step.
+
+- 2026-10-08: Made a local index error while recording a directory or
+  symlink stop a backup under `--skip-errors`
+  ([issue #284](https://git.eeqj.de/sneak/vaultik/issues/284)). Phase 2
+  only records such an entry in the local index, since it has no data
+  to open or read, but an error doing so was skipped like an unreadable
+  file. The snapshot then completed without the entry, and a restore
+  did not recreate it. The error now stops the backup, as it does
+  without the flag.
+
 - 2026-10-08: Counted a file that a backup could not store as failed
   ([issue #280](https://git.eeqj.de/sneak/vaultik/issues/280)). A file
   that phase 1 counted and phase 2 could not open, because it was

internal/cli/snapshot.go

@@ -255,8 +255,12 @@ manifest (the destination store may hold snapshots this host doesn't
 know about), which is what 'vaultik prune' does. On success the command
 prints the exact 'vaultik prune' invocation to run as a follow-up.
 
+An ID that names no snapshot in the local index or on the destination
+store is an error.
+
 Use --local-only to skip the remote half (e.g. when you want to forget a
-snapshot locally without touching the destination store).
+snapshot locally without touching the destination store); the snapshot
+must then be in the local index.
 
 If the remote is unreachable, the local-database removal still completes
 and a warning is emitted; run 'vaultik snapshot remove <snapshot-id>' again

internal/snapshot/scanner.go

@@ -1312,6 +1312,13 @@ func (s *Scanner) processPhase(
 
 	// Process each file
 	for _, fileToProcess := range filesToProcess {
+		// Check context cancellation
+		select {
+		case <-ctx.Done():
+			return ctx.Err()
+		default:
+		}
+
 		// Update progress
 		if s.progress != nil {
 			s.progress.GetStats().CurrentFile.Store(fileToProcess.Path)
@@ -1348,13 +1355,32 @@ func (s *Scanner) processPhase(
 	return s.finalizeProcessPhase(ctx, result)
 }
 
-// processFileWithErrorHandling wraps processFileStreaming with error recovery for
+// processFileWithErrorHandling records a directory or symlink, or wraps
-// deleted files and skip-errors mode. Returns (skipped, error).
+// processFileStreaming for a regular file with error recovery for deleted
+// files and skip-errors mode. Returns (skipped, error).
 func (s *Scanner) processFileWithErrorHandling(
 	ctx context.Context, fileToProcess *FileToProcess, result *ScanResult,
 ) (bool, error) {
+	// A directory or symlink has no data to open or read; it is only
+	// recorded in the local index. An error recording it stops the run
+	// even under --skip-errors, or the snapshot would complete without it.
+	mode := os.FileMode(fileToProcess.File.Mode)
+	if mode&os.ModeSymlink != 0 || mode.IsDir() {
+		err := s.recordNonRegularFile(ctx, fileToProcess)
+		if err != nil {
+			return false, fmt.Errorf("processing file %s: %w", fileToProcess.Path, err)
+		}
+
+		return false, nil
+	}
+
 	err := s.processFileStreaming(ctx, fileToProcess, result)
 	if err != nil {
+		// A cancelled run stops here even under --skip-errors, rather than
+		// counting this file as failed and going on to the next one.
+		if ctx.Err() != nil {
+			return false, fmt.Errorf("processing file %s: %w", fileToProcess.Path, err)
+		}
 		// A packer/database/encryption/upload failure means the chunk's data
 		// may not have been stored. Skipping the file would let the snapshot
 		// record a file whose chunk is in no blob and cannot be restored, so
@@ -1392,12 +1418,7 @@ func (s *Scanner) processFileWithErrorHandling(
 
 // countFailedFile counts a file that phase 2 could not store as failed
 // and takes its size back out of BytesScanned, where phase 1 put it.
-// Phase 1 counts no directories, so a directory is not counted here.
 func countFailedFile(fileToProcess *FileToProcess, result *ScanResult) {
-	if fileToProcess.FileInfo.IsDir() {
-		return
-	}
-
 	result.FilesFailed++
 	result.BytesScanned -= fileToProcess.FileInfo.Size()
 }
@@ -1770,16 +1791,11 @@ func (e *packerError) Error() string { return e.err.Error() }
 
 func (e *packerError) Unwrap() error { return e.err }
 
-// processFileStreaming processes a file by streaming chunks directly to the packer
+// processFileStreaming processes a regular file by streaming chunks directly
+// to the packer
 func (s *Scanner) processFileStreaming(
 	ctx context.Context, fileToProcess *FileToProcess, result *ScanResult,
 ) error {
-	// Symlinks and directories have no data to chunk — just record them in the DB.
-	mode := os.FileMode(fileToProcess.File.Mode)
-	if mode&os.ModeSymlink != 0 || mode.IsDir() {
-		return s.recordNonRegularFile(ctx, fileToProcess)
-	}
-
 	file, err := s.fs.Open(fileToProcess.Path)
 	if err != nil {
 		return fmt.Errorf("opening file: %w", wrapPermissionError(fileToProcess.Path, err))

internal/snapshot/skip_errors_test.go

@@ -82,6 +82,32 @@ func (f *readFailFs) Open(name string) (afero.File, error) {
 	return file, nil
 }
 
+// cancelOnOpenFs cancels the run when the scanner opens the target file to
+// back it up, as Ctrl-C partway through processing would, and records each
+// file opened after that.
+type cancelOnOpenFs struct {
+	afero.Fs
+
+	target            string
+	cancel            context.CancelFunc
+	cancelled         bool
+	openedAfterCancel []string
+}
+
+//nolint:ireturn // afero.Fs.Open is defined to return the interface.
+func (f *cancelOnOpenFs) Open(name string) (afero.File, error) {
+	if f.cancelled {
+		f.openedAfterCancel = append(f.openedAfterCancel, name)
+	}
+
+	if name == f.target {
+		f.cancel()
+		f.cancelled = true
+	}
+
+	return f.Fs.Open(name)
+}
+
 // linkRemovedAfterLstatFs is the real filesystem, except that the symlink at
 // target is removed right after the walk lstats it, as happens when a link is
 // deleted during a backup. The scanner's readlink of it then fails.
@@ -128,15 +154,16 @@ func writeSkipErrorTestFile(t *testing.T, fs afero.Fs, path, content string) {
 	}
 }
 
-// runSkipErrorScan scans source on fs with the given skip-errors setting,
+// runSkipErrorScan scans source on fs under ctx with the given skip-errors
-// printing user-facing messages to uiw (nil discards them), and returns the
+// setting, printing user-facing messages to uiw (nil discards them), and
-// repositories (for inspection) and the scan error.
+// returns the repositories (for inspection) and the scan error.
 func runSkipErrorScan(
-	t *testing.T, fs afero.Fs, source string, skipErrors bool, uiw *ui.Writer,
+	ctx context.Context, t *testing.T, fs afero.Fs, source string,
+	skipErrors bool, uiw *ui.Writer,
 ) (*database.Repositories, error) {
 	t.Helper()
 
-	db, err := database.NewTestDB()
+	db, err := database.New(ctx, ":memory:")
 	if err != nil {
 		t.Fatalf("create test db: %v", err)
 	}
@@ -161,7 +188,6 @@ func runSkipErrorScan(
 		SkipErrors:       skipErrors,
 	})
 
-	ctx := context.Background()
 	snapshotID := "test-snapshot-skip-errors"
 	createTestSnapshotRecord(ctx, t, repos, snapshotID)
 
@@ -185,7 +211,7 @@ func TestScannerPackingFailureAbortsUnderSkipErrors(t *testing.T) {
 	writeSkipErrorTestFile(t, fs, "/source/file1.txt", "first file content")
 	writeSkipErrorTestFile(t, fs, "/source/file2.txt", "second file content")
 
-	repos, err := runSkipErrorScan(t, fs, "/source", true, nil)
+	repos, err := runSkipErrorScan(context.Background(), t, fs, "/source", true, nil)
 	if err == nil {
 		t.Fatal("expected scan to abort on the packer error, got nil")
 	}
@@ -212,7 +238,7 @@ func TestScannerReadErrorAbortsWithoutSkipErrors(t *testing.T) {
 	fs := &readFailFs{Fs: afero.NewMemMapFs(), target: target}
 	writeSkipErrorTestFile(t, fs, target, "content that cannot be read")
 
-	_, err := runSkipErrorScan(t, fs, "/source", false, nil)
+	_, err := runSkipErrorScan(context.Background(), t, fs, "/source", false, nil)
 	if err == nil {
 		t.Fatal("expected scan to fail on the read error, got nil")
 	}
@@ -228,7 +254,7 @@ func TestScannerReadErrorSkippedWithSkipErrors(t *testing.T) {
 	fs := &readFailFs{Fs: afero.NewMemMapFs(), target: target}
 	writeSkipErrorTestFile(t, fs, target, "content that cannot be read")
 
-	repos, err := runSkipErrorScan(t, fs, "/source", true, nil)
+	repos, err := runSkipErrorScan(context.Background(), t, fs, "/source", true, nil)
 	if err != nil {
 		t.Fatalf("expected scan to complete with --skip-errors, got %v", err)
 	}
@@ -267,7 +293,7 @@ func TestScannerUnreadableSymlinkAbortsWithoutSkipErrors(t *testing.T) {
 	sourceDir, linkPath := writeSymlinkSource(t)
 	fs := &linkRemovedAfterLstatFs{t: t, target: linkPath}
 
-	_, err := runSkipErrorScan(t, fs, sourceDir, false, nil)
+	_, err := runSkipErrorScan(context.Background(), t, fs, sourceDir, false, nil)
 	if !errors.Is(err, os.ErrNotExist) {
 		t.Fatalf("expected scan to fail on the removed symlink, got %v", err)
 	}
@@ -283,7 +309,7 @@ func TestScannerUnreadableSymlinkSkippedWithSkipErrors(t *testing.T) {
 	fs := &linkRemovedAfterLstatFs{t: t, target: linkPath}
 	uiw := ui.NewWithColor(io.Discard, false)
 
-	repos, err := runSkipErrorScan(t, fs, sourceDir, true, uiw)
+	repos, err := runSkipErrorScan(context.Background(), t, fs, sourceDir, true, uiw)
 	if err != nil {
 		t.Fatalf("expected scan to complete with --skip-errors, got %v", err)
 	}
@@ -302,3 +328,56 @@ func TestScannerUnreadableSymlinkSkippedWithSkipErrors(t *testing.T) {
 		t.Fatalf("expected %s not to be recorded", linkPath)
 	}
 }
+
+// TestScannerCancelStopsSkipErrorsRun checks that a --skip-errors backup
+// cancelled partway through processing returns the cancellation error,
+// opens no further file, and reports no file as failed.
+func TestScannerCancelStopsSkipErrorsRun(t *testing.T) {
+	t.Parallel()
+
+	tests := []struct {
+		name          string
+		targetContent string
+	}{
+		// The cancellation lands while the target is being read.
+		{name: "while reading a file", targetContent: "first file content"},
+		// An empty target has no chunks, so the cancellation goes unnoticed
+		// until the run moves on to the next file.
+		{name: "between files", targetContent: ""},
+	}
+
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			t.Parallel()
+
+			const target = "/source/a.txt"
+
+			ctx, cancel := context.WithCancel(context.Background())
+			defer cancel()
+
+			fs := &cancelOnOpenFs{
+				Fs: afero.NewMemMapFs(), target: target, cancel: cancel,
+			}
+			writeSkipErrorTestFile(t, fs, target, tt.targetContent)
+			writeSkipErrorTestFile(t, fs, "/source/b.txt", "second file content")
+			writeSkipErrorTestFile(t, fs, "/source/c.txt", "third file content")
+
+			uiw := ui.NewWithColor(io.Discard, false)
+
+			_, err := runSkipErrorScan(ctx, t, fs, "/source", true, uiw)
+			if !errors.Is(err, context.Canceled) {
+				t.Fatalf("expected the cancellation error, got %v", err)
+			}
+
+			if len(fs.openedAfterCancel) != 0 {
+				t.Fatalf("expected no file opened after the cancellation, got %v",
+					fs.openedAfterCancel)
+			}
+
+			if uiw.ErrorCount() != 0 {
+				t.Fatalf("expected no file reported as failed, got %d error lines",
+					uiw.ErrorCount())
+			}
+		})
+	}
+}

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
+// success while storing nothing, reads that return corrupt or truncated
-// truncated bytes. It is the seam called for by the fault-injection
+// bytes, and a delete the backend refuses. It is the seam called for by
-// tests (sneak/vaultik issue 72) and is meant to be reused by future
+// the fault-injection tests (sneak/vaultik issue 72) and is meant to be
-// tests rather than re-implemented per case.
+// reused by future tests rather than re-implemented per case.
 //
-// The wrapper delegates every method to the inner Storer. Two hooks
+// The wrapper delegates every method to the inner Storer. Three hooks
-// change that: OnPut decides the fate of each write, and OnGet decides
+// change that: OnPut decides the fate of each write, OnGet decides how
-// how each read's bytes are returned. Both are keyed by the object key,
+// each read's bytes are returned, and OnDelete decides whether each
-// so a test can fault only blobs, only metadata, or a single object.
+// 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)
 }
 
@@ -171,6 +185,11 @@ func (f *Storer) ListStream(
 	return f.inner.ListStream(ctx, prefix)
 }
 
+// DeletePartialUploads delegates unchanged.
+func (f *Storer) DeletePartialUploads(ctx context.Context, prefix string) error {
+	return f.inner.DeletePartialUploads(ctx, prefix)
+}
+
 // Info delegates unchanged.
 func (f *Storer) Info() storage.Info {
 	return f.inner.Info()

internal/storage/file.go

@@ -242,6 +242,42 @@ func (f *FileStorer) ListStream(ctx context.Context, prefix string) <-chan Objec
 	return ch
 }
 
+// DeletePartialUploads removes every file under prefix whose name ends in
+// tempSuffix. A missing prefix has none to remove.
+func (f *FileStorer) DeletePartialUploads(ctx context.Context, prefix string) error {
+	basePath := f.fullPath(prefix)
+
+	exists, err := afero.Exists(f.fs, basePath)
+	if err != nil {
+		return fmt.Errorf("checking path: %w", err)
+	}
+
+	if !exists {
+		return nil
+	}
+
+	err = afero.Walk(f.fs, basePath, func(path string, info os.FileInfo, err error) error {
+		if err != nil {
+			return err
+		}
+
+		if ctx.Err() != nil {
+			return ctx.Err()
+		}
+
+		if info.IsDir() || !strings.HasSuffix(info.Name(), tempSuffix) {
+			return nil
+		}
+
+		return f.fs.Remove(path)
+	})
+	if err != nil {
+		return fmt.Errorf("walking directory: %w", err)
+	}
+
+	return nil
+}
+
 // Info returns human-readable storage location information.
 func (f *FileStorer) Info() Info {
 	return Info{

internal/storage/file_atomic_test.go

@@ -120,3 +120,45 @@ func TestFileStorer_ListSkipsPartialFiles(t *testing.T) {
 		t.Fatalf("ListStream should return only the real key, got %v", streamed)
 	}
 }
+
+// TestFileStorer_DeletePartialUploads checks that a leftover temp file is
+// removed and the object at the real key is kept.
+func TestFileStorer_DeletePartialUploads(t *testing.T) {
+	t.Parallel()
+
+	base := t.TempDir()
+
+	f, err := storage.NewFileStorer(base)
+	if err != nil {
+		t.Fatalf("NewFileStorer: %v", err)
+	}
+
+	ctx := context.Background()
+
+	err = f.Put(ctx, testBlobKey, strings.NewReader("blob-bytes"))
+	if err != nil {
+		t.Fatalf("Put: %v", err)
+	}
+
+	leftover := filepath.Join(base, testBlobKey+"-123456.partial")
+
+	err = os.WriteFile(leftover, []byte("half"), 0o600)
+	if err != nil {
+		t.Fatalf("writing leftover temp file: %v", err)
+	}
+
+	err = f.DeletePartialUploads(ctx, "blobs/")
+	if err != nil {
+		t.Fatalf("DeletePartialUploads: %v", err)
+	}
+
+	_, err = os.Stat(leftover)
+	if !os.IsNotExist(err) {
+		t.Errorf("leftover temp file was not removed: %v", err)
+	}
+
+	_, err = f.Stat(ctx, testBlobKey)
+	if err != nil {
+		t.Errorf("Stat of the real key: %v", err)
+	}
+}

internal/storage/rclone.go

@@ -213,6 +213,31 @@ func (r *RcloneStorer) ListStream(
 	return ch
 }
 
+// DeletePartialUploads removes every object under prefix whose name ends
+// in tempSuffix.
+func (r *RcloneStorer) DeletePartialUploads(ctx context.Context, prefix string) error {
+	var partial []fs.Object
+
+	err := operations.ListFn(ctx, r.fsys, func(obj fs.Object) {
+		key := obj.Remote()
+		if strings.HasPrefix(key, prefix) && strings.HasSuffix(key, tempSuffix) {
+			partial = append(partial, obj)
+		}
+	})
+	if err != nil {
+		return fmt.Errorf("listing objects: %w", err)
+	}
+
+	for _, obj := range partial {
+		err = obj.Remove(ctx)
+		if err != nil {
+			return fmt.Errorf("removing object: %w", err)
+		}
+	}
+
+	return nil
+}
+
 // Info returns human-readable storage location information.
 func (r *RcloneStorer) Info() Info {
 	location := r.remote

internal/storage/rclone_test.go

@@ -198,6 +198,47 @@ func TestRcloneStorerListSkipsPartialFiles(t *testing.T) {
 	}
 }
 
+// TestRcloneStorerDeletePartialUploads checks that a temporary file left
+// by a killed upload is removed and the object at the real key is kept.
+//
+//nolint:paralleltest // NewRcloneStorer installs the process-global rclone config
+func TestRcloneStorerDeletePartialUploads(t *testing.T) {
+	dir := t.TempDir()
+	ctx := context.Background()
+
+	s, err := storage.NewRcloneStorer(ctx, ":local", dir)
+	if err != nil {
+		t.Fatalf("NewRcloneStorer: %v", err)
+	}
+
+	err = s.Put(ctx, testBlobKey, strings.NewReader("blob-bytes"))
+	if err != nil {
+		t.Fatalf("Put: %v", err)
+	}
+
+	leftover := filepath.Join(dir, testBlobKey+"-123456.partial")
+
+	err = os.WriteFile(leftover, []byte("half"), 0o600)
+	if err != nil {
+		t.Fatalf("writing leftover temp file: %v", err)
+	}
+
+	err = s.DeletePartialUploads(ctx, "blobs/")
+	if err != nil {
+		t.Fatalf("DeletePartialUploads: %v", err)
+	}
+
+	_, err = os.Stat(leftover)
+	if !os.IsNotExist(err) {
+		t.Errorf("leftover temp file was not removed: %v", err)
+	}
+
+	_, err = s.Stat(ctx, testBlobKey)
+	if err != nil {
+		t.Errorf("Stat of the real key: %v", err)
+	}
+}
+
 // newRcloneStorerOnWrappedLocal registers name as rclone's local backend
 // wrapped by wrap, and builds an rclone backend on it rooted at a fresh
 // temp directory. wrap changes the features the local backend reports, so

internal/storage/s3.go

@@ -99,6 +99,12 @@ func (s *S3Storer) ListStream(ctx context.Context, prefix string) <-chan ObjectI
 	return ch
 }
 
+// DeletePartialUploads has nothing to remove: S3 shows an object only once
+// its upload has completed, so an upload cut off part-way leaves none.
+func (s *S3Storer) DeletePartialUploads(_ context.Context, _ string) error {
+	return nil
+}
+
 // Info returns human-readable storage location information.
 func (s *S3Storer) Info() Info {
 	return Info{

internal/storage/storer.go

@@ -71,6 +71,12 @@ type Storer interface {
 	// If an error occurs during listing, the final item will have Err set.
 	ListStream(ctx context.Context, prefix string) <-chan ObjectInfo
 
+	// DeletePartialUploads removes every object under prefix that an
+	// upload cut off part-way left under a temporary name ending in
+	// `.partial`. The file and rclone backends' List and ListStream skip
+	// such an object.
+	DeletePartialUploads(ctx context.Context, prefix string) error
+
 	// Info returns human-readable storage location information.
 	Info() Info
 }

internal/vaultik/destination_validation_test.go

@@ -257,6 +257,10 @@ func (s *stubLister) List(_ context.Context, _ string) ([]string, error) {
 	return nil, errStubUnused
 }
 
+func (s *stubLister) DeletePartialUploads(_ context.Context, _ string) error {
+	return errStubUnused
+}
+
 func (s *stubLister) Info() storage.Info {
 	return storage.Info{}
 }

internal/vaultik/helpers.go

@@ -3,6 +3,7 @@ package vaultik
 import (
 	"errors"
 	"fmt"
+	"math"
 	"regexp"
 	"strconv"
 	"strings"
@@ -47,6 +48,10 @@ const (
 	year  = 365 * day
 )
 
+// maxDuration is the largest value a time.Duration can hold, about 292
+// years.
+const maxDuration = time.Duration(math.MaxInt64)
+
 // A snapshot ID split on "_" has at least a hostname and a trailing
 // timestamp.
 const minSnapshotIDParts = 2
@@ -145,10 +150,10 @@ func parseSnapshotName(snapshotID, hostname string) string {
 // parseDuration parses a duration string with support for human-friendly units:
 // d/day/days, w/week/weeks, mo/month/months, y/year/years, plus standard Go
 // duration units. Following Go, m is minutes and mo is months. A bare number,
-// an unknown unit, and a negative value are all rejected. Outside the Go
+// an unknown unit, a negative value, and a total longer than maxDuration are
-// units the input must be whole numbers each followed directly by a unit
+// all rejected. Outside the Go units the input must be whole numbers each
-// (2w3d), so 1.0y and 30 days are errors; decimals work only in Go units
+// followed directly by a unit (2w3d), so 1.0y and 30 days are errors;
-// (1.5h).
+// decimals work only in Go units (1.5h).
 func parseDuration(s string) (time.Duration, error) {
 	if strings.HasPrefix(strings.TrimSpace(s), "-") {
 		return 0, errNegativeDuration
@@ -181,19 +186,30 @@ func parseDuration(s string) (time.Duration, error) {
 			return 0, fmt.Errorf("invalid number %q: %w", match[1], err)
 		}
 
+		var unitLength time.Duration
+
 		unit := strings.ToLower(match[2])
 		switch unit {
 		case "d", "day", "days":
-			total += time.Duration(n) * day
+			unitLength = day
 		case "w", "week", "weeks":
-			total += time.Duration(n) * week
+			unitLength = week
 		case "mo", "month", "months":
-			total += time.Duration(n) * month
+			unitLength = month
 		case "y", "year", "years":
-			total += time.Duration(n) * year
+			unitLength = year
 		default:
 			return 0, fmt.Errorf("%w %q", errUnknownTimeUnit, unit)
 		}
+
+		// Past maxDuration the sum wraps to a negative value, which would
+		// put the purge cutoff in the future.
+		if time.Duration(n) > (maxDuration-total)/unitLength {
+			return 0, fmt.Errorf("%w: %q is longer than about 292 years",
+				errInvalidDuration, s)
+		}
+
+		total += time.Duration(n) * unitLength
 	}
 
 	return total, nil

internal/vaultik/helpers_test.go

@@ -111,6 +111,17 @@ func TestParseDuration(t *testing.T) {
 		{"30 days", 0, true},
 		{"30 days ago", 0, true},
 		{"x7d", 0, true},
+		// The largest time.Duration is about 292 years. A larger value
+		// must be rejected, not wrapped to a negative one, which would
+		// put the cutoff in the future and select every snapshot.
+		{"292y", 292 * 365 * 24 * time.Hour, false},
+		{"293y", 0, true},
+		{"3559mo", 0, true},
+		{"15251w", 0, true},
+		{"106752d", 0, true},
+		{"2562048h", 0, true},
+		{"200y200y", 0, true},
+		{"292y6mo", 0, true},
 	}
 
 	for _, tt := range tests {

internal/vaultik/integration_test.go

@@ -155,6 +155,12 @@ func (m *MockStorer) ListStream(
 	return ch
 }
 
+// DeletePartialUploads has nothing to remove: Put stores each object
+// under its key at once.
+func (m *MockStorer) DeletePartialUploads(_ context.Context, _ string) error {
+	return nil
+}
+
 func (m *MockStorer) Info() storage.Info {
 	return storage.Info{
 		Type:     "mock",

internal/vaultik/nuke_remote_test.go

@@ -0,0 +1,144 @@
+package vaultik_test
+
+import (
+	"context"
+	"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
+// under a file:// destination, including the `.partial` files uploads
+// killed part-way leave next to a snapshot's metadata and next to where a
+// blob would have been.
+func TestNukeRemoteLeavesNoFiles(t *testing.T) {
+	log.Initialize(log.Config{})
+	t.Parallel()
+
+	ctx := context.Background()
+	storeDir := filepath.Join(t.TempDir(), "store")
+	v, repos, out := backUpToFileDestination(ctx, t, storeDir)
+
+	snapshots, err := repos.Snapshots.ListRecent(ctx, listRecentTestLimit)
+	require.NoError(t, err)
+	require.Len(t, snapshots, 1)
+
+	snapshotKey := snapshot.RemoteSnapshotKey(snapshots[0].ID.String())
+	hash := testBlobHashA
+	leftovers := []string{
+		filepath.Join(storeDir, "metadata", snapshotKey,
+			"db.zst.age-123456.partial"),
+		filepath.Join(storeDir, "blobs", hash[:2], hash[2:4],
+			hash+"-123456.partial"),
+	}
+
+	for _, leftover := range leftovers {
+		require.NoError(t, os.MkdirAll(filepath.Dir(leftover), 0o750))
+		require.NoError(t, os.WriteFile(leftover, []byte("half an upload"), 0o600))
+	}
+
+	require.NoError(t, v.NukeRemote(true))
+
+	var files []string
+
+	err = filepath.WalkDir(storeDir,
+		func(path string, entry fs.DirEntry, err error) error {
+			if err != nil {
+				return err
+			}
+
+			if !entry.IsDir() {
+				files = append(files, path)
+			}
+
+			return nil
+		})
+	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")
+		})
+	}
+}
+
+// TestNukeRemoteRefusesAnotherDestinationWithoutRetryAdvice checks that
+// remote nuke, when the local index belongs to a different destination
+// than the configured one, fails without advising to run nuke again:
+// that run would fail the same way.
+func TestNukeRemoteRefusesAnotherDestinationWithoutRetryAdvice(t *testing.T) {
+	log.Initialize(log.Config{})
+	t.Parallel()
+
+	ctx := context.Background()
+	storeDir := filepath.Join(t.TempDir(), "store")
+	v, _, out := backUpToFileDestination(ctx, t, storeDir)
+
+	v.Config.StorageURL = "file://" + filepath.Join(t.TempDir(), "elsewhere")
+
+	err := v.NukeRemote(true)
+
+	require.ErrorContains(t, err, "bound to a different backup destination")
+	assert.NotContains(t, out.String(), "vaultik remote nuke --force")
+}
+
+// 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

@@ -19,13 +19,22 @@ 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"
 
 // NukeRemote deletes every snapshot's metadata and every blob from remote
-// storage. After this returns successfully the bucket prefix is empty and
+// storage, along with any object an upload cut off part-way left under a
-// the next backup starts from scratch.
+// temporary `.partial` name. After this returns successfully the bucket
+// prefix is empty and the next backup starts from scratch.
 //
 // Refuses to run unless force is true. The caller is responsible for
 // confirming with the user.
@@ -34,6 +43,29 @@ func (v *Vaultik) NukeRemote(force bool) error {
 		return errNukeRequiresForce
 	}
 
+	// Checked before anything is deleted: running nuke again would fail
+	// the same way, so the advice below must not follow this error.
+	err := v.EnsureStorageBinding()
+	if err != nil {
+		return err
+	}
+
+	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})
@@ -48,7 +80,14 @@ func (v *Vaultik) NukeRemote(force bool) error {
 		return fmt.Errorf("pruning blobs: %w", err)
 	}
 
-	v.UI.Completef("Backup destination store is now empty.")
+	// The file and rclone listings skip `.partial` objects, so the two
+	// steps above never delete them.
+	for _, prefix := range []string{"metadata/", "blobs/"} {
+		err = v.Storage.DeletePartialUploads(v.ctx, prefix)
+		if err != nil {
+			return fmt.Errorf("deleting partial uploads: %w", err)
+		}
+	}
 
 	return nil
 }
@@ -152,14 +191,17 @@ 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))
+	}
 
 	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.ErrorContains(t, err, "1 blob(s) could not be deleted")
+	assert.FileExists(t, orphan)
+}

internal/vaultik/purge_per_name_test.go

@@ -267,6 +267,29 @@ func TestPurgeOlderThan_WithNameFilter(t *testing.T) {
 	assert.Contains(t, remaining, snapHomeT0)
 }
 
+func TestPurgeOlderThan_TooLargeDeletesNothing(t *testing.T) {
+	log.Initialize(log.Config{})
+	t.Parallel()
+
+	snapshotIDs := []string{snapHomeT0, snapHomeT1}
+
+	v := setupPurgeTest(t, snapHostname, snapshotIDs)
+
+	err := v.PurgeSnapshotsWithOptions(&vaultik.SnapshotPurgeOptions{
+		OlderThan: "1000y",
+		Force:     true,
+	})
+	require.Error(t, err)
+
+	err = v.PurgeSnapshotsWithOptions(&vaultik.SnapshotPurgeOptions{
+		OlderThan: "292y",
+		Force:     true,
+	})
+	require.NoError(t, err)
+
+	assert.ElementsMatch(t, snapshotIDs, listRemainingSnapshots(t, v))
+}
+
 func TestPurgeKeepLatest_ThreeNames(t *testing.T) {
 	log.Initialize(log.Config{})
 	t.Parallel()

internal/vaultik/remove_snapshot_test.go

@@ -3,17 +3,21 @@ package vaultik_test
 import (
 	"bytes"
 	"context"
+	"database/sql"
 	"io"
 	"strings"
 	"sync"
 	"testing"
+	"time"
 
 	"github.com/klauspost/compress/zstd"
 	"github.com/stretchr/testify/assert"
 	"github.com/stretchr/testify/require"
+	"sneak.berlin/go/vaultik/internal/database"
 	"sneak.berlin/go/vaultik/internal/log"
 	"sneak.berlin/go/vaultik/internal/snapshot"
 	"sneak.berlin/go/vaultik/internal/storage"
+	"sneak.berlin/go/vaultik/internal/types"
 	"sneak.berlin/go/vaultik/internal/vaultik"
 )
 
@@ -125,6 +129,12 @@ func (s *testStorer) ListStream(
 	return ch
 }
 
+// DeletePartialUploads has nothing to remove: Put stores each object
+// under its key at once.
+func (s *testStorer) DeletePartialUploads(_ context.Context, _ string) error {
+	return nil
+}
+
 func (s *testStorer) Info() storage.Info {
 	return storage.Info{
 		Type:     testLabel,
@@ -208,6 +218,37 @@ func addBlob(t *testing.T, store *testStorer, hash string) {
 	require.NoError(t, err)
 }
 
+// newLocalIndex returns an in-memory local index holding a completed
+// snapshot row for each of ids.
+func newLocalIndex(t *testing.T, ids ...string) *database.Repositories {
+	t.Helper()
+
+	ctx := context.Background()
+
+	db, err := database.New(ctx, ":memory:")
+	require.NoError(t, err)
+	t.Cleanup(func() { _ = db.Close() })
+
+	repos := database.NewRepositories(db)
+
+	for _, id := range ids {
+		completedAt := time.Now().UTC()
+		snap := &database.Snapshot{
+			ID:          types.SnapshotID(id),
+			Hostname:    snapHostname,
+			StartedAt:   completedAt,
+			CompletedAt: &completedAt,
+		}
+
+		err = repos.WithTx(ctx, func(ctx context.Context, tx *sql.Tx) error {
+			return repos.Snapshots.Create(ctx, tx, snap)
+		})
+		require.NoError(t, err)
+	}
+
+	return repos
+}
+
 // ============================================================================
 // Unit Tests for RemoveSnapshot
 // ============================================================================
@@ -227,6 +268,7 @@ func TestRemoveSnapshot_LocalOnly_PreservesRemote(t *testing.T) {
 	addBlob(t, store, blobA)
 
 	tv := vaultik.NewForTesting(store)
+	tv.Repositories = newLocalIndex(t, "snapshot-001")
 
 	opts := &vaultik.RemoveOptions{Force: true, LocalOnly: true}
 	result, err := tv.RemoveSnapshot("snapshot-001", opts)
@@ -262,6 +304,7 @@ func TestRemoveSnapshot_DefaultRemovesMetadataNotBlobs(t *testing.T) {
 	addBlob(t, store, blobShared)
 
 	tv := vaultik.NewForTesting(store)
+	tv.Repositories = newLocalIndex(t, "snapshot-001", "snapshot-002")
 
 	opts := &vaultik.RemoveOptions{Force: true}
 	result, err := tv.RemoveSnapshot("snapshot-001", opts)
@@ -310,6 +353,84 @@ func TestRemoveSnapshot_DryRun(t *testing.T) {
 	assert.Contains(t, tv.Stdout.String(), "[Dry run - no changes made]")
 }
 
+// TestRemoveSnapshot_UnknownIDIsError checks that an ID naming no
+// snapshot in the local index or on the destination store is an error
+// under every option, and that nothing is reported as removed.
+func TestRemoveSnapshot_UnknownIDIsError(t *testing.T) {
+	log.Initialize(log.Config{})
+	t.Parallel()
+
+	for name, opts := range map[string]vaultik.RemoveOptions{
+		"force":      {Force: true},
+		"json":       {JSON: true},
+		"dry-run":    {Force: true, DryRun: true},
+		"local-only": {Force: true, LocalOnly: true},
+	} {
+		t.Run(name, func(t *testing.T) {
+			t.Parallel()
+
+			store := newTestStorer()
+			addManifest(t, store, "snapshot-001", nil)
+
+			tv := vaultik.NewForTesting(store)
+			tv.Repositories = newLocalIndex(t, "snapshot-001")
+
+			result, err := tv.RemoveSnapshot("snapshot-999", &opts)
+
+			require.ErrorContains(t, err, "snapshot not found")
+			assert.False(t, result.RemoteRemoved)
+			assert.Empty(t, tv.Stdout.String())
+			assert.True(t, store.hasKey(remoteKeyPath("snapshot-001")))
+		})
+	}
+}
+
+// TestRemoveSnapshot_NoMetadataOnDestination checks that removing a
+// snapshot whose metadata is not on the destination store does not
+// report metadata removed from it.
+func TestRemoveSnapshot_NoMetadataOnDestination(t *testing.T) {
+	log.Initialize(log.Config{})
+	t.Parallel()
+
+	tv := vaultik.NewForTesting(newTestStorer())
+	tv.Repositories = newLocalIndex(t, "snapshot-001")
+
+	result, err := tv.RemoveSnapshot("snapshot-001",
+		&vaultik.RemoveOptions{Force: true})
+
+	require.NoError(t, err)
+	assert.False(t, result.RemoteRemoved)
+
+	out := tv.Stdout.String()
+	assert.Contains(t, out, "Removed snapshot 'snapshot-001' from local database")
+	assert.NotContains(t, out, "Removed snapshot metadata from remote storage")
+}
+
+// TestRemoveSnapshot_NotInLocalIndex checks that removing a snapshot
+// only the destination store holds removes its metadata there without
+// reporting a removal from the local database.
+func TestRemoveSnapshot_NotInLocalIndex(t *testing.T) {
+	log.Initialize(log.Config{})
+	t.Parallel()
+
+	store := newTestStorer()
+	addManifest(t, store, "snapshot-001", nil)
+
+	tv := vaultik.NewForTesting(store)
+	tv.Repositories = newLocalIndex(t)
+
+	result, err := tv.RemoveSnapshot("snapshot-001",
+		&vaultik.RemoveOptions{Force: true})
+
+	require.NoError(t, err)
+	assert.True(t, result.RemoteRemoved)
+	assert.False(t, store.hasKey(remoteKeyPath("snapshot-001")))
+
+	out := tv.Stdout.String()
+	assert.NotContains(t, out, "from local database")
+	assert.Contains(t, out, "Removed snapshot metadata from remote storage")
+}
+
 func TestRemoveAllSnapshots_RequiresForce(t *testing.T) {
 	log.Initialize(log.Config{})
 	t.Parallel()
@@ -355,7 +476,8 @@ 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 +23,8 @@ var (
 	errNoSnapshotsInConfig  = errors.New("no snapshots configured")
 	errSnapshotVerifyFailed = errors.New("verification failed")
 	errRemoveAllNeedsForce  = errors.New("--all requires --force")
+	errSnapshotNotFound     = errors.New("snapshot not found")
+	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 " +
@@ -643,7 +645,7 @@ func (v *Vaultik) confirmAndExecutePurge(
 				"snapshot_id", snapshotID, "error", err)
 		}
 
-		err = v.deleteRemoteSnapshotByKey(snapshot.RemoteSnapshotKey(snapshotID))
+		_, err = v.deleteRemoteSnapshotByKey(snapshot.RemoteSnapshotKey(snapshotID))
 		if err != nil {
 			return fmt.Errorf("deleting snapshot %s from remote: %w", snapshotID, err)
 		}
@@ -1153,6 +1155,11 @@ func (v *Vaultik) RemoveSnapshot(
 		return result, err
 	}
 
+	inLocalIndex, err := v.findSnapshotToRemove(snapshotID, opts)
+	if err != nil {
+		return result, err
+	}
+
 	if opts.DryRun {
 		return v.removeSnapshotDryRun(result, snapshotID, opts)
 	}
@@ -1161,12 +1168,14 @@ func (v *Vaultik) RemoveSnapshot(
 		return result, nil
 	}
 
+	if inLocalIndex {
 		log.Info("Removing snapshot from local database", "snapshot_id", snapshotID)
 
 		err = v.deleteSnapshotFromLocalDB(snapshotID)
 		if err != nil {
 			return result, fmt.Errorf("removing from local database: %w", err)
 		}
+	}
 
 	if !opts.LocalOnly {
 		result.RemoteRemoved = v.removeSnapshotRemote(snapshotID, opts)
@@ -1184,9 +1193,11 @@ func (v *Vaultik) RemoveSnapshot(
 		return result, v.outputRemoveJSON(result)
 	}
 
+	if inLocalIndex {
 		v.stdoutf("Removed snapshot '%s' from local database\n", snapshotID)
+	}
 
-	if !opts.LocalOnly && result.RemoteRemoved {
+	if result.RemoteRemoved {
 		v.printlnStdout("Removed snapshot metadata from remote storage")
 		v.stdoutf("\nNote: The removed snapshot's blobs remain on the remote. "+
 			"Run '%s' to delete any blobs no longer referenced by any "+
@@ -1196,6 +1207,45 @@ func (v *Vaultik) RemoveSnapshot(
 	return result, nil
 }
 
+// findSnapshotToRemove reports whether the local index has a row for
+// snapshotID, and returns errSnapshotNotFound when there is nothing to
+// remove: no row and, unless LocalOnly is set, no metadata on the
+// destination store either. The destination store is consulted only
+// when the local index has no row, so a snapshot in the local index can
+// still be removed from it while the destination store is unreachable.
+func (v *Vaultik) findSnapshotToRemove(
+	snapshotID string, opts *RemoveOptions,
+) (bool, error) {
+	if v.Repositories != nil {
+		snap, err := v.Repositories.Snapshots.GetByID(v.ctx, snapshotID)
+		if err != nil {
+			return false, fmt.Errorf("looking up snapshot in local index: %w", err)
+		}
+
+		if snap != nil {
+			return true, nil
+		}
+	}
+
+	if opts.LocalOnly {
+		return false, fmt.Errorf("%w in the local index: %s",
+			errSnapshotNotFound, snapshotID)
+	}
+
+	objects, err := v.remoteSnapshotObjects(snapshot.RemoteSnapshotKey(snapshotID))
+	if err != nil {
+		return false, fmt.Errorf("looking up snapshot on destination store: %w", err)
+	}
+
+	if len(objects) == 0 {
+		return false, fmt.Errorf(
+			"%w in the local index or on the destination store: %s",
+			errSnapshotNotFound, snapshotID)
+	}
+
+	return false, nil
+}
+
 // removeSnapshotDryRun reports what RemoveSnapshot would do without
 // making changes.
 func (v *Vaultik) removeSnapshotDryRun(
@@ -1255,14 +1305,14 @@ func (v *Vaultik) confirmRemoveSnapshot(snapshotID string, opts *RemoveOptions)
 // didn't finish and to run `vaultik snapshot remove` for the snapshot
 // again once the destination store is reachable (`vaultik prune` never
 // removes snapshot metadata). Returns true when the remote removal
-// succeeded.
+// succeeded and deleted something.
 func (v *Vaultik) removeSnapshotRemote(snapshotID string, opts *RemoveOptions) bool {
 	log.Info("Removing snapshot metadata from remote storage",
 		"snapshot_id", snapshotID)
 
 	remoteKey := snapshot.RemoteSnapshotKey(snapshotID)
 
-	err := v.deleteRemoteSnapshotByKey(remoteKey)
+	deleted, err := v.deleteRemoteSnapshotByKey(remoteKey)
 	if err != nil {
 		log.Warn("Could not remove snapshot metadata from remote storage; "+
 			"run '"+snapshotRemoveCommandHint+"' with the snapshot's ID "+
@@ -1280,7 +1330,7 @@ func (v *Vaultik) removeSnapshotRemote(snapshotID string, opts *RemoveOptions) b
 		return false
 	}
 
-	return true
+	return deleted > 0
 }
 
 // RemoveAllSnapshots removes every snapshot known to the local
@@ -1464,8 +1514,8 @@ 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
+// orphan remote keys. Blobs are not touched. It returns an error when
-// hint is the next step.
+// any snapshot could not be removed.
 func (v *Vaultik) executeRemoveAll(
 	localSnaps, orphanRemoteKeys []string, opts *RemoveOptions,
 ) (*RemoveResult, error) {
@@ -1479,18 +1529,11 @@ 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)
+		failures += v.removeAllOrphanRemotes(orphanRemoteKeys, result)
-
+		result.RemoteRemoved = failures == 0
-		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)
-		}
 	}
 
 	if v.SnapshotManager != nil {
@@ -1500,6 +1543,10 @@ 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 +1555,6 @@ 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 +1597,13 @@ 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
+// --local-only, its remote metadata), returning the number of snapshots
-// deletion failures.
+// 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)
@@ -1568,24 +1613,26 @@ func (v *Vaultik) removeAllLocalSnapshots(
 			log.Error("Failed to remove from local database",
 				"snapshot_id", snapshotID, "error", err)
 
+			failures++
+
 			continue
 		}
 
 		if !opts.LocalOnly {
-			err = v.deleteRemoteSnapshotByKey(
+			_, err = v.deleteRemoteSnapshotByKey(
 				snapshot.RemoteSnapshotKey(snapshotID))
 			if err != nil {
 				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
@@ -1598,7 +1645,7 @@ func (v *Vaultik) removeAllOrphanRemotes(
 	for _, key := range orphanRemoteKeys {
 		log.Info("Removing orphan remote snapshot", "remote_key", key)
 
-		err := v.deleteRemoteSnapshotByKey(key)
+		_, err := v.deleteRemoteSnapshotByKey(key)
 		if err != nil {
 			log.Warn("Failed to remove orphan from remote",
 				"remote_key", key, "error", err)
@@ -1645,34 +1692,44 @@ func (v *Vaultik) deleteSnapshotFromLocalDB(snapshotID string) error {
 }
 
 // deleteRemoteSnapshotByKey removes everything under
-// metadata/<remoteKey>/ on the destination store. The argument is a
+// metadata/<remoteKey>/ on the destination store and returns how many
-// remote key (double-SHA256 derivation), not a human snapshot ID;
+// objects it deleted. The argument is a remote key (double-SHA256
-// callers that have a human ID must hash via snapshot.RemoteSnapshotKey
+// derivation), not a human snapshot ID; callers that have a human ID
-// first.
+// must hash via snapshot.RemoteSnapshotKey first.
-func (v *Vaultik) deleteRemoteSnapshotByKey(remoteKey string) error {
+func (v *Vaultik) deleteRemoteSnapshotByKey(remoteKey string) (int, error) {
-	prefix := fmt.Sprintf("metadata/%s/", remoteKey)
+	objectsToDelete, err := v.remoteSnapshotObjects(remoteKey)
-	objectCh := v.Storage.ListStream(v.ctx, prefix)
-
-	var objectsToDelete []string
-
-	for object := range objectCh {
-		if object.Err != nil {
-			return fmt.Errorf("listing objects: %w", object.Err)
-		}
-
-		objectsToDelete = append(objectsToDelete, object.Key)
-	}
-
-	for _, key := range objectsToDelete {
-		err := v.Storage.Delete(v.ctx, key)
 	if err != nil {
-			return fmt.Errorf("removing %s: %w", key, err)
+		return 0, err
+	}
+
+	for i, key := range objectsToDelete {
+		err = v.Storage.Delete(v.ctx, key)
+		if err != nil {
+			return i, fmt.Errorf("removing %s: %w", key, err)
 		}
 
 		log.Debug("Deleted remote object", "key", key)
 	}
 
-	return nil
+	return len(objectsToDelete), nil
+}
+
+// remoteSnapshotObjects lists the keys of every object under
+// metadata/<remoteKey>/ on the destination store.
+func (v *Vaultik) remoteSnapshotObjects(remoteKey string) ([]string, error) {
+	prefix := fmt.Sprintf("metadata/%s/", remoteKey)
+
+	var keys []string
+
+	for object := range v.Storage.ListStream(v.ctx, prefix) {
+		if object.Err != nil {
+			return nil, fmt.Errorf("listing objects: %w", object.Err)
+		}
+
+		keys = append(keys, object.Key)
+	}
+
+	return keys, nil
 }
 
 // outputRemoveJSON outputs the removal result as JSON

internal/vaultik/snapshot_skip_errors_test.go

@@ -0,0 +1,75 @@
+package vaultik_test
+
+import (
+	"context"
+	"fmt"
+	"path/filepath"
+	"testing"
+
+	"github.com/spf13/afero"
+	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
+	"sneak.berlin/go/vaultik/internal/config"
+	"sneak.berlin/go/vaultik/internal/database"
+	"sneak.berlin/go/vaultik/internal/log"
+	"sneak.berlin/go/vaultik/internal/storage"
+	"sneak.berlin/go/vaultik/internal/vaultik"
+)
+
+// --skip-errors skips only a file that cannot be opened or read. A local
+// index error while a directory is recorded stops the backup, and the
+// snapshot is not recorded as complete. See
+// https://git.eeqj.de/sneak/vaultik/issues/284.
+func TestSkipErrorsBackupStopsOnIndexErrorRecordingDirectory(t *testing.T) {
+	log.Initialize(log.Config{})
+	t.Parallel()
+
+	ctx := context.Background()
+
+	// The scan walks the source path with symlinks resolved, so dirPath
+	// must be spelled the same way to match the row the scan inserts.
+	tempDir, err := filepath.EvalSymlinks(t.TempDir())
+	require.NoError(t, err)
+
+	srcDir := filepath.Join(tempDir, "src")
+	dirPath := filepath.Join(srcDir, "dir")
+
+	fs := afero.NewOsFs()
+	require.NoError(t, fs.MkdirAll(dirPath, 0o755))
+	require.NoError(t, afero.WriteFile(fs,
+		filepath.Join(dirPath, "file.txt"), []byte("file content"), 0o644))
+
+	cfg := faultTestConfig()
+	cfg.IndexPath = filepath.Join(tempDir, "index.sqlite")
+	cfg.Snapshots = map[string]config.SnapshotConfig{
+		"tree": {Paths: []string{srcDir}},
+	}
+
+	store, err := storage.NewFileStorer(filepath.Join(tempDir, "remote"))
+	require.NoError(t, err)
+
+	db, err := database.New(ctx, cfg.IndexPath)
+	require.NoError(t, err)
+	t.Cleanup(func() { _ = db.Close() })
+
+	// The local index refuses the directory's files row.
+	_, err = db.Conn().ExecContext(ctx, fmt.Sprintf(`
+		CREATE TRIGGER refuse_directory BEFORE INSERT ON files
+		WHEN NEW.path = '%s'
+		BEGIN SELECT RAISE(ABORT, 'simulated index error'); END`, dirPath))
+	require.NoError(t, err)
+
+	repos := database.NewRepositories(db)
+	v := newBackupVaultik(ctx, cfg, store, repos, db, fs)
+
+	err = v.CreateSnapshot(&vaultik.SnapshotCreateOptions{
+		SkipErrors: true,
+		Snapshots:  []string{"tree"},
+	})
+	require.ErrorContains(t, err, "simulated index error")
+
+	snapshots, err := repos.Snapshots.ListRecent(ctx, listRecentTestLimit)
+	require.NoError(t, err)
+	require.Len(t, snapshots, 1)
+	assert.Nil(t, snapshots[0].CompletedAt)
+}

internal/vaultik/storage_bind.go

@@ -9,7 +9,7 @@ import (
 )
 
 // errStorageBindingMismatch is returned when the local index database is
-// bound to a different destination than the configured storage_url.
+// bound to a different destination than the configured one.
 var errStorageBindingMismatch = errors.New(
 	"local index is bound to a different backup destination")
 
@@ -24,8 +24,9 @@ var errStorageBindingMismatch = errors.New(
 //
 // Behaviour:
 //   - On first use (empty stored value), record the configured
-//     storage_url and log the binding.
+//     destination (storage_url, or the s3.* endpoint, bucket and
-//   - When the stored value matches the configured storage_url, do
+//     prefix) and log the binding.
+//   - When the stored value matches the configured destination, do
 //     nothing and return nil.
 //   - When the two differ, refuse with an error that tells the user
 //     how to recover (revert the config, or run `vaultik database
@@ -46,12 +47,10 @@ func (v *Vaultik) EnsureStorageBinding() error {
 
 	configured := v.Config.StorageURL
 	if configured == "" {
-		// Some legacy configs still use the split s3.* keys instead of
+		// The s3.* form: bind on its endpoint, bucket and prefix,
-		// storage_url. Falling back to a synthetic URL for those would
+		// written the way storage_url would name the same destination.
-		// hide the fact that the binding is loose. Instead, treat
+		s3 := v.Config.S3
-		// unset as "nothing to bind against" — the check is
+		configured = "s3://" + s3.Bucket + "/" + s3.Prefix + "?endpoint=" + s3.Endpoint
-		// necessarily best-effort for those configs.
-		return nil
 	}
 
 	stored, err := v.Repositories.LocalMeta.Get(v.ctx, database.LocalMetaKeyStorageURL)
@@ -96,7 +95,8 @@ func buildBindingMismatchMessage(stored, configured string) string {
 		"future snapshots referencing blobs that don't exist at the new destination.\n" +
 		"\n" +
 		"To proceed, either:\n" +
-		"  - revert storage_url in your config to the bound destination, or\n" +
+		"  - revert storage_url (or the s3.* endpoint, bucket and prefix) in your\n" +
+		"    config to the bound destination, or\n" +
 		"  - run 'vaultik database delete' to discard the local index and rebuild it\n" +
 		"    from a fresh full backup against the new destination"
 }

internal/vaultik/storage_bind_test.go

@@ -76,3 +76,51 @@ func TestEnsureStorageBinding_MismatchRefuses(t *testing.T) {
 	assert.Contains(t, err.Error(), "file:///mnt/backups/new")
 	assert.Contains(t, err.Error(), "vaultik database delete")
 }
+
+// A config with no storage_url names its destination with s3.endpoint,
+// s3.bucket and s3.prefix; a change to any of them must be refused just
+// as a changed storage_url is.
+func TestEnsureStorageBinding_S3FieldsChangeRefuses(t *testing.T) {
+	log.Initialize(log.Config{})
+	t.Parallel()
+
+	bound := config.S3Config{
+		Endpoint: "https://s3.example.com",
+		Bucket:   "bucket-a",
+		Prefix:   "backups",
+	}
+
+	tests := []struct {
+		name    string
+		changed config.S3Config
+	}{
+		{"bucket", config.S3Config{
+			Endpoint: bound.Endpoint, Bucket: "bucket-b", Prefix: bound.Prefix,
+		}},
+		{"prefix", config.S3Config{
+			Endpoint: bound.Endpoint, Bucket: bound.Bucket, Prefix: "other",
+		}},
+		{"endpoint", config.S3Config{
+			Endpoint: "https://s3.other.com", Bucket: bound.Bucket, Prefix: bound.Prefix,
+		}},
+	}
+
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			t.Parallel()
+
+			tv, _ := buildBindTestVaultik(t, "")
+
+			tv.Config.S3 = bound
+			require.NoError(t, tv.EnsureStorageBinding(), "first use binds")
+			require.NoError(t, tv.EnsureStorageBinding(), "same destination passes")
+
+			tv.Config.S3 = tt.changed
+			err := tv.EnsureStorageBinding()
+			require.Error(t, err)
+			assert.Contains(t, err.Error(),
+				"local index is bound to a different backup destination")
+			assert.Contains(t, err.Error(), "vaultik database delete")
+		})
+	}
+}