weaviate/weaviate · error

shard %s of class %s has no assigned nodes

Error message

shard %s of class %s has no assigned nodes

What it means

ShardOwnership validates that every physical shard in the class's sharding state has at least one node in BelongsToNodes. A shard with an empty BelongsToNodes list has no owner, so shards cannot be assigned to nodes for a balanced export; the read fails immediately with this error.

Source

Thrown at adapters/repos/db/export.go:56

// are write-locked simultaneously. Short enough that the added latency
// (vs. event-driven wake-up) is negligible compared to typical lock hold
// times (shard load/unload), long enough to keep CPU overhead trivial.
const lockPollInterval = 5 * time.Millisecond

// ShardOwnership returns a map of node name to shard names for a given class.
// Shards are distributed across their replica nodes using a least-loaded
// strategy so that export work is balanced across the cluster.
func (db *DB) ShardOwnership(ctx context.Context, className string) (map[string][]string, error) {
	shardNodes := make(map[string][]string)

	err := db.schemaReader.Read(className, true, func(_ *models.Class, state *sharding.State) error {
		if state == nil {
			return fmt.Errorf("unable to retrieve sharding state for class %s", className)
		}

		for shardName, shard := range state.Physical {
			if len(shard.BelongsToNodes) == 0 {
				return fmt.Errorf("shard %s of class %s has no assigned nodes", shardName, className)
			}

			// Filter out empty node names to avoid assigning shards to an invalid node.
			validNodes := make([]string, 0, len(shard.BelongsToNodes))
			for _, node := range shard.BelongsToNodes {
				if node != "" {
					validNodes = append(validNodes, node)
				}
			}
			if len(validNodes) == 0 {
				return fmt.Errorf("shard %s of class %s has only empty assigned nodes", shardName, className)
			}

			shardNodes[shardName] = validNodes
		}

		return nil
	})

View on GitHub (pinned to 75aa4b6d11)

Solutions

1. Inspect the sharding state (nodes/replication status endpoints) to find the unassigned shard
2. Run the shard replication/rebalance mechanism to assign nodes to the shard
3. If the shard is orphaned (no data needed), delete it so the sharding state is consistent
4. Restore a consistent sharding state from a cluster snapshot if assignment metadata is lost

// Inspect sharding state before export; fail fast on unassigned shards
err := db.schemaReader.Read(className, true, func(_ *models.Class, st *sharding.State) error {
    for name, shard := range st.Physical {
        if len(shard.BelongsToNodes) == 0 {
            return fmt.Errorf("shard %s unassigned; fix before export", name)
        }
    }
    return nil
})

_, err := db.ShardOwnership(ctx, className)
if err != nil && strings.Contains(err.Error(), "has no assigned nodes") {
    // trigger replication/rebalance or delete the orphaned shard before exporting
    return err
}

When it happens

Trigger: Calling ShardOwnership when the class's sharding.State.Physical map contains a shard whose BelongsToNodes slice is empty/nil — e.g. shard metadata created but replica assignment never committed (interrupted shard creation, corrupted sharding state).

Common situations: Crashed or interrupted shard creation/migration in a cluster; restoring a partial backup; shard rebalancing that left a shard unassigned; exporting a class right after adding shards that have not been fully assigned.

Related errors

• failed to get shard ownership for class %s: %w
• prepare request: %w
• prepare failed: status %d (%s)
• commit request: %w
• commit failed: status %d (%s)

AI-assisted analysis of weaviate/weaviate@75aa4b6d11 (2026-09-04). Data as JSON: /api/errors/d86c380b2f4451d5. Report an issue: GitHub.