What causes "error fetching posting list" in dgraph-io/dgraph?

Inspect the underlying Badger error logged alongside this sentinel to find the root cause

error fetching posting list

Error message

error fetching posting list

What it means

errFetchingPostingList is a sentinel error returned by HNSW helper functions (getDataFromKeyWithCacheType, populateEdgeDataFromKeyWithCacheType, getVecFromUid, removeDeadNodes, SearchWithUid/Options) when reading the posting list for a key from Badger fails (the inner error is typically wrapped/logged). It means the vector-index layer could not fetch the underlying postings data for a UID or key.

Source

Thrown at tok/hnsw/helper.go:51

	visitedVectorsLevel  = "visited_vectors_level_"
	distanceComputations = "vector_distance_computations"
	searchTime           = "vector_search_time"
	VecEntry             = "__vector_entry"
	VecDead              = "__vector_dead"
	VectorIndexMaxLevels = 5
	EfConstruction       = 16
	EfSearch             = 12
	// ByteData indicates the key stores data.
	ByteData = byte(0x00)
	// DefaultPrefix is the prefix used for data, index and reverse keys so that relative
	DefaultPrefix = byte(0x00)
	// NsSeparator is the separator between the namespace and attribute.
	NsSeparator = "-"
)

var (
	errNilVector           = errors.New("nil vector returned")
	errFetchingPostingList = errors.New("error fetching posting list")
)

type SearchResult struct {
	nnUids        []uint64
	traversalPath []uint64
	extraMetrics  map[string]uint64
}

func (s *SearchResult) GetNnUids() []uint64 {
	return s.nnUids
}

func (s *SearchResult) GetTraversalPath() []uint64 {
	return s.traversalPath
}

func (s *SearchResult) GetExtraMetrics() map[string]uint64 {
	return s.extraMetrics

View on GitHub (pinned to 759e242be6)

Solutions

Inspect the underlying Badger error logged alongside this sentinel to find the root cause
Retry the query; transient IO/lock issues may resolve
Run Badger integrity checks / backup-restore tooling if corruption is suspected
Rebuild the vector index for the predicate if postings remain unreadable

Example fix

// before
res, err := hnsw.SearchWithUid(ctx, index, uid, opts) // errFetchingPostingList
// after
if errors.Is(err, tok.ErrFetchingPostingList) {
    log.Warnf("posting list read failed for uid %d, retrying", uid)
    res, err = hnsw.SearchWithUid(ctx, index, uid, opts)
}

Defensive patterns

Strategy: retry

Validate before calling

// health check before vector queries
if err := dbHealthCheck(ctx); err != nil {
    return fmt.Errorf("postings store unhealthy: %w", err)
}

Try / catch

res, err := hnsw.SearchWithUid(ctx, idx, uid, opts)
if err != nil {
    if errors.Is(err, errFetchingPostingList) {
        return backoffRetry(ctx, 3, func() error {
            r, e := hnsw.SearchWithUid(ctx, idx, uid, opts); res = r; return e
        })
    }
    return err
}

Prevention

Monitor Badger disk health and IO errors; check logs for the wrapped root cause
Run integrity/backup tooling after crashes or restores
Retry transient read failures with backoff before surfacing to users
Rebuild vector indexes whose postings stay unreadable

When it happens

Trigger: Vector search (SearchWithUid, SearchWithUidAndOptions) or index maintenance (removeDeadNodes) hitting a posting-list read error: DB I/O failure, key corruption, or key not readable at the given read timestamp.

Common situations: Badger disk/IO issues or corrupted SST files after crash; querying with a readTs below/at a problematic timestamp; restore/migration leaving inconsistent postings for vector predicates.

Related errors

AI-assisted analysis of dgraph-io/dgraph@759e242be6 (2026-09-01). Data as JSON: /api/errors/3ceda2d9054252a8. Report an issue: GitHub.