What causes "add meta count" in weaviate/weaviate?

Check the wrapped cause: 'objects bucket is nil' means the shard store isn't open — restart the node or inspect shard directory health

weaviate/weaviate · error

add meta count

Error message

add meta count

What it means

Wrapper error produced by unfilteredAggregator.Do when IncludeMetaCount is set and addMetaCount fails. addMetaCount counts all objects in the shard's objects bucket; it fails if the bucket is missing or the underlying LSM Count call errors. The wrap only adds context — the root cause is in the wrapped error.

Source

Thrown at adapters/repos/db/aggregator/unfiltered.go:49

// However, this aggregator does not work with subselections of the dataset,
// such as when grouping or a filter is set.
type unfilteredAggregator struct {
	*Aggregator
}

func newUnfilteredAggregator(agg *Aggregator) *unfilteredAggregator {
	return &unfilteredAggregator{Aggregator: agg}
}

func (ua *unfilteredAggregator) Do(ctx context.Context) (*aggregation.Result, error) {
	out := aggregation.Result{}

	// without grouping there is always exactly one group
	out.Groups = make([]aggregation.Group, 1)

	if ua.params.IncludeMetaCount {
		if err := ua.addMetaCount(ctx, &out); err != nil {
			return nil, errors.Wrap(err, "add meta count")
		}
	}

	props, err := ua.properties(ctx)
	if err != nil {
		return nil, errors.Wrap(err, "aggregate properties")
	}

	out.Groups[0].Properties = props

	return &out, nil
}

func (ua *unfilteredAggregator) addMetaCount(ctx context.Context,
	out *aggregation.Result,
) error {
	b := ua.store.Bucket(helpers.ObjectsBucketLSM)
	if b == nil {

View on GitHub (pinned to 75aa4b6d11)

Solutions

Check the wrapped cause: 'objects bucket is nil' means the shard store isn't open — restart the node or inspect shard directory health
If the cause is 'count objects', check for context cancellation/timeout and retry with a longer deadline
Verify the collection/shard exists and the node finished startup before issuing meta-count aggregations
Check disk health and LSM store logs for corruption; restore from backup if the shard is damaged

Example fix

// before (client)
res, err := graphql.Batch(aggWithMetaCount) // fails with 'add meta count: objects bucket is nil'
// after
if shardReady(node, class, shard) { res, err := graphql.Batch(aggWithMetaCount) }

Defensive patterns

Strategy: try-catch

Validate before calling

// client: ensure collection + node ready before meta-count
status, _ := client.Schema().GetSchemaStatus(ctx, className)
if status == nil || status.Shards == nil { return errors.New("collection not ready") }

Type guard

func shardReady(b BucketGetter) bool { return b != nil && b.Bucket(helpers.ObjectsBucketLSM) != nil }

Try / catch

res, err := agg.Do(ctx)
if err != nil {
  if strings.Contains(err.Error(), "objects bucket is nil") {
    // shard store not open: restart/inspect shard, then retry
  } else if errors.Is(err, context.Canceled) || errors.Is(err, context.DeadlineExceeded) {
    // retry with longer deadline
  }
  return err
}

Prevention

Verify shard health after node startup before serving aggregations
Set generous timeouts for meta-count on large collections
Monitor LSM store open errors in startup logs
Keep backups so torn shard dirs can be restored

When it happens

Trigger: A GraphQL { Meta { count } } or aggregate-with-metaCount query against an unfiltered aggregation where the shard's objects bucket is nil (shard still initializing, corrupted, or closed) or b.Count(ctx) returns an LSM/ctx error.

Common situations: Querying a shard while it is being loaded or dropped; corrupted/partially-created shard directories after a crash; context canceled or timed out during a long count on a large collection.

Related errors

AI-assisted analysis of weaviate/weaviate@75aa4b6d11 (2026-09-04). Data as JSON: /api/errors/a0fccfa215e35401. Report an issue: GitHub.