Failed to load committed snapshot: omitting sequence number…

Error message

Failed to load committed snapshot: omitting sequence number from notifications

What it means

After committing a snapshot, MergingSnapshotProducer notifies listeners using the just-saved snapshot. If the snapshot cannot be found in the (refreshed) table metadata — typically because the latest metadata couldn't be loaded due to eventual-consistency problems during refresh — this WARN is logged, the sequence number is set to INVALID_SEQUENCE_NUMBER, and notifications are emitted without it.

Solutions

Use a catalog/storage backend with strong read-after-write consistency so the committed snapshot is immediately visible on refresh.
Retry the refresh or add a short delay before reading metadata in code that triggers notification right after commit.
If you consume notifications downstream, tolerate a null/invalid sequence number rather than assuming it is always set.
Upgrade the catalog implementation if known eventual-consistency behavior is causing frequent occurrences.

Example fix

// before (listener assumes sequence number always present)
long seq = notification.sequenceNumber();
// after
long seq = notification.sequenceNumber() != TableMetadata.INVALID_SEQUENCE_NUMBER
    ? notification.sequenceNumber()
    : -1L; // handle missing sequence number

Defensive patterns

Strategy: try-catch

Validate before calling

// after commit, confirm the snapshot is visible before notifying
if (table.operations().current().snapshot(snapshotId) == null) {
  // metadata not yet visible (eventual consistency) — handle missing sequence number
}

Type guard

static boolean hasSequenceNumber(Snapshot s) {
  return s != null && s.sequenceNumber() != TableMetadata.INVALID_SEQUENCE_NUMBER;
}

Try / catch

try {
  long seq = justSaved.sequenceNumber();
} catch (NullPointerException | IllegalStateException e) {
  long seq = TableMetadata.INVALID_SEQUENCE_NUMBER; // degrade notifications
}

Prevention

Use strongly consistent catalogs/stores so committed snapshots are visible immediately.
Design downstream notification consumers to tolerate invalid/absent sequence numbers.
Add retry/delay around refresh-after-commit in custom commit flows.
Track how often this WARN fires as an eventual-consistency signal for your storage.

When it happens

Trigger: A commit succeeds on the catalog, then table.refresh()/loading the current metadata fails to observe the new snapshot immediately (eventually consistent catalog or object store); notify listeners path runs with justSaved == null.

Common situations: Eventually consistent object stores (e.g. S3 without strong read-after-write) backing a Hadoop/FileIO catalog; slow catalog propagation right after commit; listeners (e.g. Spark listener buses, event hooks) observing notifications missing sequence numbers.

Related errors

AI-assisted analysis of apache/iceberg@86d9c8fc54 (2026-09-12). Data as JSON: /api/errors/3f14e7c20b7bd026. Report an issue: GitHub.

Appendix: source

Thrown at core/src/main/java/org/apache/iceberg/MergingSnapshotProducer.java:1101

    return manifests;
  }

  @Override
  public Object updateEvent() {
    long snapshotId = snapshotId();

    Snapshot justSaved = ops().current().snapshot(snapshotId);
    if (justSaved == null) {
      justSaved = ops().refresh().snapshot(snapshotId);
    }

    long sequenceNumber = TableMetadata.INVALID_SEQUENCE_NUMBER;
    Map<String, String> summary;
    if (justSaved == null) {
      // The snapshot just saved may not be present if the latest metadata couldn't be loaded due to
      // eventual
      // consistency problems in refresh.
      LOG.warn("Failed to load committed snapshot: omitting sequence number from notifications");
      summary = summary();
    } else {
      sequenceNumber = justSaved.sequenceNumber();
      summary = justSaved.summary();
    }

    return new CreateSnapshotEvent(tableName, operation(), snapshotId, sequenceNumber, summary);
  }

  @Override
  protected void cleanUncommitted(Set<ManifestFile> committed) {
    mergeManager.cleanUncommitted(committed);
    filterManager.cleanUncommitted(committed);
    deleteMergeManager.cleanUncommitted(committed);
    deleteFilterManager.cleanUncommitted(committed);
    cleanUncommittedAppends(committed);
  }

View on GitHub (pinned to 86d9c8fc54)