What causes "Schema state not found for %s." in dgraph-io/dgraph?

Retry the query after the schema mutation settles; if it persists, restart/rebalance so nodes reload a consistent schema.

dgraph-io/dgraph · error

Schema state not found for %s.

Error message

Schema state not found for %s.

What it means

After confirming the attribute is indexed, pickTokenizer calls schema.State().Tokenizer to fetch the tokenizer list; if it comes back nil despite the attribute being indexed, this error is thrown. It indicates the in-memory schema state lacks tokenizer metadata for a predicate that IsIndexed claims is indexed.

Source

Thrown at worker/tokens.go:86

	if funcType == ngramFn {
		if query {
			return tok.GetNGramQueryTokens(funcArgs, lang)
		} else {
			return tok.GetNGramTokens(funcArgs, lang)
		}
	}
	return tok.GetTermTokens(funcArgs)
}

func pickTokenizer(ctx context.Context, attr string, f string) (tok.Tokenizer, error) {
	// Get the tokenizers and choose the corresponding one.
	if !schema.State().IsIndexed(ctx, attr) {
		return nil, errors.Errorf("Attribute %s is not indexed.", attr)
	}

	tokenizers := schema.State().Tokenizer(ctx, attr)
	if tokenizers == nil {
		return nil, errors.Errorf("Schema state not found for %s.", attr)
	}
	for _, t := range tokenizers {
		// If function is eq and we found a tokenizer that's !Lossy(), lets return it
		switch f {
		case "eq":
			// For equality, find a non-lossy tokenizer.
			if !t.IsLossy() {
				return t, nil
			}
		default:
			// rest of the cases: ge, gt, le, lt require a sortable tokenizer.
			if t.IsSortable() {
				return t, nil
			}
		}
	}

	// Should we return an error if we don't find a non-lossy tokenizer for eq function.

View on GitHub (pinned to 759e242be6)

Solutions

Retry the query after the schema mutation settles; if it persists, restart/rebalance so nodes reload a consistent schema.
Re-apply the schema alteration (drop and re-add the index) so tokenizer metadata is rebuilt.
Check cluster schema consistency (all nodes report the same schema) and fix stale nodes.

Example fix

// before: predicate altered while queries in flight
// after: re-apply schema and retry
await dgraph.Alter(ctx, "name: string @index(exact) .")
// then run the eq query

Defensive patterns

Strategy: retry

Try / catch

err := runQuery(ctx)
if err != nil && strings.Contains(err.Error(), "Schema state not found") {
    time.Sleep(backoff)
    err = runQuery(ctx) // retry after schema state settles
}

Prevention

Avoid running queries concurrently with schema alterations where possible.
Ensure all cluster nodes have converged on the same schema after mutations.
Re-apply index alterations if tokenizer metadata appears lost.

When it happens

Trigger: An indexed predicate whose schema entry has no tokenizer entry — typically a race or inconsistency between schema state and the tokenizer registry, e.g. querying during/just after a schema mutation, or a predicate indexed in one way that lost its tokenizer metadata.

Common situations: Concurrent schema alterations while queries run; cluster where a node has stale/partial schema state; predicates with index types that do not map to string tokenizers (note: for non-string indexed predicates hitting compare functions, the not-indexed or no-proper-index errors are more typical).

Related errors

AI-assisted analysis of dgraph-io/dgraph@759e242be6 (2026-09-01). Data as JSON: /api/errors/dc3d0273b80b6c32. Report an issue: GitHub.