Commits & concurrency

4 h
requires LKH-01

From your world. Overlapping dbt runs writing the same table

Practice quiz with instant feedback. For a scored run that counts toward your local progress, use uv run de quiz LKH-02.

1What does Iceberg's optimistic concurrency control assume, and how does a commit detect a conflict? AIt locks the table for the whole write, so no conflict is possible. BIt assumes the base version will not change. Then it commits with a compare-and-swap that fails if the table moved off the base version. CIt merges all concurrent writes automatically, and a failure is not possible. DIt serializes writers with a queue in the catalog.

2Two writers both read balance=100. Each writer reads the row and then overwrites it with a new balance. The first writer commits 110. The second writer (still based on 100) commits 105. What happened? AFinal balance is 115, the correct serial result. BA lost update: the +10 of the first writer is erased, and the final balance is 105. CThe catalog rejects both writes. DThe balances are summed to 215.

3After a commit raises a conflict (CommitFailedException), what is the correct recovery? ACatch the exception and ignore it. BImmediately resubmit the exact same precomputed metadata without reloading. CRefresh the table to the state of the winner, recompute the change on that fresh base, and retry the commit (bounded retries). DDelete the table and start over.

4Which kind of operation can Iceberg most often re-apply automatically after a conflict, without you recomputing? AA read-modify-write overwrite of a row based on a value you read. BA pure append of new files (it has no requirement on existing files). CA delete of specific files that another writer already removed. DA schema change racing another schema change.

6How does refresh-and-reapply guarantee no lost update across two conflicting read-modify-write operations? AIt does not. One update is always lost. BThe retried operation recomputes its change on the committed (fresh) state. So the final result equals running the two operations serially. CIt locks the table during the retry. DIt sums both results.