apache/iceberg · error · RuntimeException

Interrupted during commit

Error message

Interrupted during commit

What it means

If the commit thread is interrupted while waiting or communicating with HMS, doCommit re-interrupts the thread and throws a RuntimeException 'Interrupted during commit'. The commit outcome may be unknown, so callers must check the table state before retrying.

Solutions

1. Allow the commit to complete before cancelling the job; avoid killing tasks mid-commit
2. After interruption, verify the table's metadata_location to see if the commit landed before retrying
3. Increase timeouts so commits aren't interrupted by supervising frameworks
4. Re-run the operation on a fresh (refreshed) table snapshot if the commit did not land

Example fix

// before
future.cancel(true); // interrupts an in-flight commit
// after
future.get(10, TimeUnit.MINUTES); // wait for commit completion before teardown

// avoid committing inside interruptible cancellation scopes; check Thread.interrupted() before starting a long commit

try {
  table.newAppend().appendFile(f).commit();
} catch (RuntimeException e) {
  if (e.getMessage() != null && e.getMessage().contains("Interrupted during commit")) {
    verifyCommitLanded(catalog.loadTable(ident)); // check metadata_location before retry
  } else throw e;
}

When it happens

Trigger: doCommit interrupted via Thread.interrupt() — e.g. task cancellation in Spark/Flink, executor shutdown, or query timeout killing the writing task while it is inside the HMS commit path.

Common situations: Spark stage cancellation/kill; Flink job cancel; application shutdown hooks; speculative-execution cleanup interrupting a slow commit.

Understand the failure class

Background: Request timed out: what client-side request timeouts mean across libraries (Request timed out, TIMED_OUT, APITimeoutError) — this error's family across 39 libraries.

Related errors

• Cannot commit: Base metadata location
• Interrupted during commit
• Interrupted during refresh
• Interrupted during refresh
• Interrupted in call to check table existence of

AI-assisted analysis of apache/iceberg@86d9c8fc54 (2026-09-12). Data as JSON: /api/errors/1ec696aaecb3712a. Report an issue: GitHub.

Appendix: source

Thrown at hive-metastore/src/main/java/org/apache/iceberg/hive/HiveTableOperations.java:420

          commitStatus = checkCommitStatus(newMetadataLocation, tableMetadata);
        }

        switch (commitStatus) {
          case SUCCESS:
            break;
          case FAILURE:
            throw e;
          case UNKNOWN:
            throw new CommitStateUnknownException(e);
        }
      }
    } catch (TException e) {
      throw new RuntimeException(
          String.format("Metastore operation failed for %s.%s", database, tableName), e);

    } catch (InterruptedException e) {
      Thread.currentThread().interrupt();
      throw new RuntimeException("Interrupted during commit", e);

    } catch (LockException e) {
      throw new CommitFailedException(e);

    } finally {
      HiveOperationsBase.cleanupMetadataAndUnlock(io(), commitStatus, newMetadataLocation, lock);
    }

    LOG.info(
        "Committed to table {} with the new metadata location {}", fullName, newMetadataLocation);
  }

  @Override
  public long maxHiveTablePropertySize() {
    return maxHiveTablePropertySize;
  }

  @Override

View on GitHub (pinned to 86d9c8fc54)