vitessio/vitess · error

streams are mismatched across source shards

Error message

streams are mismatched across source shards: %v

What it means

This is the complementary direction of the stream-mismatch check: after processing all rows on a shard, if any reference entries remain unmatched (`len(ref) != 0`), the shard is missing streams that other source shards have. The remaining unmatched entries are printed in the error to show exactly what is missing.

Solutions

1. Read the error output listing the missing streams (binlog source, cell, tablet types).
2. Recreate the missing streams on the deficient shard — ideally by deleting the workflow and re-running the original MoveTables/Reshard creation so all shards get identical streams.
3. If recreating via tooling, ensure the shard is properly registered and primary-serving so VReplication can be set up there.
4. Verify with `SELECT workflow, binlog_source, cell, tablet_types FROM _vt.vreplication` across all source shards before re-running the reshard.

Example fix

-- before: shard 1 has 1 stream, shard 0 has 2
-- recreate missing stream on shard 1 (or rerun MoveTables)
-- after: both shards report the same stream set

perShard := map[string]int{} // shard -> stream count for workflow
for _, shard := range sourceShards { perShard[shard] = countStreams(shard, workflow) }
if len(uniqueCounts(perShard)) > 1 { return fmt.Errorf("stream counts differ across source shards") }

func allShardsHaveStreams(perShard map[string]int) bool {
    want := maxCount(perShard)
    for _, n := range perShard { if n != want { return false } }
    return true
}

if err := wr.Reshard(...); err != nil {
    if strings.Contains(err.Error(), "streams are mismatched across source shards:") {
        // error body lists missing streams; recreate them and retry
        return recreateMissingStreams(ctx, err)
    }
    return err
}

When it happens

Trigger: Reshard run where one source shard has fewer VReplication streams for a workflow than the reference shard — the leftover `ref` set is reported with this message.

Common situations: A shard's stream crashed and was deleted and not recreated; a partial MoveTables that only created streams on some shards; manual cleanup removed a stream on one shard; a shard was added after the workflow started and never got its streams.

Related errors

• streams are mismatched across source shards for workflow
• cannot reshard streams with a mix of reference and sharded…
• Create ReshardWorkflow failed
• source shard is not in serving state
• streams are mismatched across source shards for workflow

AI-assisted analysis of vitessio/vitess@01a25a7d17 (2026-09-01). Data as JSON: /api/errors/c6d69ac674159cc4. Report an issue: GitHub.

Appendix: source

Thrown at go/vt/wrangler/resharder.go:249

				continue
			}
			key := fmt.Sprintf("%s:%s:%s", workflow, bls.Keyspace, bls.Shard)
			if mustCreate {
				rs.refStreams[key] = &refStream{
					workflow:    workflow,
					bls:         &bls,
					cell:        row[2].ToString(),
					tabletTypes: row[3].ToString(),
				}
			} else {
				if !ref[key] {
					return fmt.Errorf("streams are mismatched across source shards for workflow: %s", workflow)
				}
				delete(ref, key)
			}
		}
		if len(ref) != 0 {
			return fmt.Errorf("streams are mismatched across source shards: %v", ref)
		}
		return nil
	})
	return err
}

// blsIsReference is partially copied from streamMigrater.templatize.
// It reuses the constants from that function also.
func (rs *resharder) blsIsReference(bls *binlogdatapb.BinlogSource) (bool, error) {
	streamType := workflow.StreamTypeUnknown
	for _, rule := range bls.Filter.Rules {
		typ, err := rs.identifyRuleType(rule)
		if err != nil {
			return false, err
		}

		switch typ {
		case workflow.StreamTypeSharded:

View on GitHub (pinned to 01a25a7d17)