mjsax opened a new pull request #8886:
URL: https://github.com/apache/kafka/pull/8886
A StandbyTask must wipe its local state if EOS is enabled and no local
checkpoint file is found.
I left debug statement in on purpose to people can verify the test locally
if they want.
- when the first instance starts up, the store should be empty
- if should process key 0 and put into the store
- after the second instance is running and has replicated the state, the
first instance should process key 1 and crash
- after fail-over to instance 2, the store should only contain key 0
- instance 2 should reprocess key 1 (we skip it to avoid the crash for this
case; note that the store content in not modified; we only need to do this to
make sure instance 1 can be restarted and rebuild the standby store)
- instance 1 is restarted and rebuild the standby store
- we stop instance 2 an the store is migrated to instance 1: the store
should only contain key 0 (if the fix in `StandbyTask` is remove, one can
observe that the store content is 0 and 1)
- we re-enable error injection and re-inject the poison pill: instance 1
should fail again (without the fix, no error occurs because key 1 is detected
as duplicated)
Call for review @abbccdda @guozhangwang
----------------------------------------------------------------
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.
For queries about this service, please contact Infrastructure at:
us...@infra.apache.org
