numinnex opened a new pull request, #4738: URL: https://github.com/apache/iggy/pull/4738
Closes #4631 A crash rolls a `replicated` partition back to the last WAL frontier it published, at a checkpoint, a durable operation or a graceful shutdown, and discards what follows, even bytes already in every node's segment files. This PR takes that as the intended contract: keeping the tail would need a recovery design that handles torn writes and keeps election safety, and topics that must survive a whole-cluster crash can use `persisted`. Approving it accepts that contract. The README and the server config described the loss as only the writes "not yet handed to the OS". Both now describe the frontier rollback. The change is identical to the one in #4736, so whichever PR merges second applies cleanly. `cluster::crash_durability` gains the `replicated` twin of the eager-flush whole-cluster test, using the setup of the closed #4632 with the expectation flipped. Today none of its 30 acks survive, so the test fails if recovery starts keeping the tail without the docs and this test changing with it. The module doc no longer claims a whole-cluster crash keeps everything that reached the segments. -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected]
