WXPNG opened a new issue, #12566:
URL: https://github.com/apache/seatunnel/issues/12566
Hi,
The Zeta engine already recovers running jobs after an active-master switch
via `CoordinatorService#restoreAllRunningJobFromMasterNodeSwitch()`. We'd
like
to confirm the intended scope of this recovery, because we observed two gaps
when the **entire cluster** (all master + worker nodes) is restarted
together,
rather than a single-master failover:
1. Running jobs are sometimes silently lost. `runningJobInfoIMap` is backed
by
a Hazelcast EAGER MapStore; on a full restart the persisted entries load
asynchronously, so `entrySet()` may return empty (no exception) and the
restore path returns without restoring anything.
2. "Normal" jobs (submitted without a savepoint) re-sync from scratch instead
of resuming from the latest checkpoint, because `restoreSourceJobId` is
null
and `CheckpointManager` skips checkpoint restore in that case.
So my questions are:
- Is automatic recovery after a full cluster restart part of the official
design/roadmap? If yes, the two behaviors above look like bugs.
- If it is intentionally out of scope, is the recommended practice to rely on
manual savepoint / `--restore-with-checkpoint`, or is there a plan to
support
it?
Environment: SeaTunnel dev branch, Zeta cluster mode, Java 8/11.
Thanks!
--
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]
