weaviate/weaviate · error

failed to get version for %d

Error message

failed to get version for %d

What it means

VersionStore.Get in the hfresh vector index reads a vector's version byte from the persistent LSMKV version bucket. This wrapped error is returned when the underlying bucket.Get call itself fails (I/O error, corrupted store, closed bucket, failed bucket acquisition path in wrapped form). It is distinct from ErrVectorNotFound, which is returned when the key simply has no stored version.

Source

Thrown at adapters/repos/db/vector/hfresh/version_map.go:285

func (v *VersionStore) key(vectorID uint64) []byte {
	buf := make([]byte, len(versionMapBucketPrefix)+8)
	copy(buf, versionMapBucketPrefix)
	binary.LittleEndian.PutUint64(buf[len(versionMapBucketPrefix):], vectorID)
	return buf
}

func (v *VersionStore) Get(ctx context.Context, vectorID uint64) (VectorVersion, error) {
	key := v.key(vectorID)
	bucket, release, err := v.bucket.acquire()
	if err != nil {
		return 0, err
	}
	defer release()

	version, err := bucket.Get(key[:])
	if err != nil {
		return 0, errors.Wrapf(err, "failed to get version for %d", vectorID)
	}

	if len(version) == 0 {
		return 0, ErrVectorNotFound
	}

	return VectorVersion(version[0]), nil
}

func (v *VersionStore) Set(ctx context.Context, vectorID uint64, version VectorVersion) error {
	key := v.key(vectorID)
	bucket, release, err := v.bucket.acquire()
	if err != nil {
		return err
	}
	defer release()

	return bucket.Put(key[:], []byte{byte(version)})

View on GitHub (pinned to 75aa4b6d11)

Solutions

1. Check disk health and free space on the persistence data volume (dmesg, df).
2. Verify the shard/collection was not deleted or closed while clients were reading; retry after shard is available.
3. If the store is corrupted, restore the affected shard/collection from a backup or re-index the data.
4. Check file permissions/ownership of the Weaviate data directory (PERSISTENCE_DATA_PATH).

Example fix

// caller: distinguish 'not found' from real read failure
version, err := store.Get(ctx, vectorID)
if err != nil {
    if errors.Is(err, ErrVectorNotFound) {
        // key absent — handle as missing, do not retry
        return nil, errVectorMissing
    }
    // wrapped "failed to get version for %d" — storage-level, retry or alert
    return nil, fmt.Errorf("version store read failed: %w", err)
}

// Go: classify via errors.Is against the sentinel
func isVectorNotFound(err error) bool { return errors.Is(err, ErrVectorNotFound) }
func isStorageFailure(err error) bool {
    return err != nil && !isVectorNotFound(err)
}

version, err := store.Get(ctx, vectorID)
if err != nil {
    switch {
    case errors.Is(err, ErrVectorNotFound):
        // treat as absent vector
    default:
        // wrapped storage failure: retry with backoff / alert on persistence
        return retryOrEscalate(err)
    }
}

When it happens

Trigger: Calling VersionStore.Get(ctx, vectorID) when the LSMKV bucket cannot be read: the bucket was acquired via v.bucket.acquire() successfully but bucket.Get(key[:]) fails due to disk I/O errors, a corrupted/unmounted store, a closed or shutting-down shard, or LSMKV internal read failures.

Common situations: Disk full or failing disk under the Weaviate persistence directory; shard being dropped or closed concurrently while a read is in flight; corrupted LSMKV segment files after an unclean shutdown; permission problems on the store directory after a container/volume migration.

Related errors

• resetDimensionsLSM: no bucket dimensions
• generate unique bucket name
• replace dimensions bucket
• cleanup quarantined %d segment(s) after exhausting the retry
• read pending/quarantine set after %d consecutive errors: %w

AI-assisted analysis of weaviate/weaviate@75aa4b6d11 (2026-09-04). Data as JSON: /api/errors/e218af52940394ea. Report an issue: GitHub.