jsancio commented on PR #12274:
URL: https://github.com/apache/kafka/pull/12274#issuecomment-1213269855
Build failed with test errors. Rerunning:
```
[Build / JDK 8 and Scala 2.12 /
jsancio commented on PR #12274:
URL: https://github.com/apache/kafka/pull/12274#issuecomment-1212113291
@dengziming take a look at the tests failures. All of the failures in
`ClusterControlManagerTest` seems related to this change.
--
This is an automated message from the Apache Git
jsancio commented on PR #12274:
URL: https://github.com/apache/kafka/pull/12274#issuecomment-1206920706
@dengziming I marked this issue as a blocker to 3.3.0. Let me know if you
want to work on this issue. I don't mind.
--
This is an automated message from the Apache Git Service.
To
jsancio commented on PR #12274:
URL: https://github.com/apache/kafka/pull/12274#issuecomment-1189538831
> @jsancio I think the last question is that there may be more than one
registration record for each broker after restarting it, so can we rely on the
broker epoch? I think I need some
jsancio commented on PR #12274:
URL: https://github.com/apache/kafka/pull/12274#issuecomment-1187345352
> This is a good idea, however, the offset of each registration record is a
soft state(not persist to metadata log)
@dengziming The offset is persistent along with the record. This
jsancio commented on PR #12274:
URL: https://github.com/apache/kafka/pull/12274#issuecomment-1183190708
> maybe we should allow the lastAppliedOffset to be behind by A records (or
time) to unfence the broker.
@dengziming Here are my thoughts
### Improve logic for unfencing
jsancio commented on PR #12274:
URL: https://github.com/apache/kafka/pull/12274#issuecomment-1177949852
> I havn't flesh it, the basic idea is shown in this PR, only read up to
Isolation.COMMITTED when reading from observer and invoke
fetchPurgatory.maybeComplete(highWatermark.offset,
jsancio commented on PR #12274:
URL: https://github.com/apache/kafka/pull/12274#issuecomment-1176530495
> In this PR I tried your suggestion and it does solve this problem,
however, this will make the logic in RaftClient very complex and we need to
save more states in LeaderState and it's
jsancio commented on PR #12274:
URL: https://github.com/apache/kafka/pull/12274#issuecomment-1176512896
> > here is the code change:
https://github.com/dengziming/kafka/tree/KAFKA-13959-2
>
> Hey @dengziming, I took at look at the commits in this tree. Is this the
only commit
jsancio commented on PR #12274:
URL: https://github.com/apache/kafka/pull/12274#issuecomment-1176510158
> here is the code change:
https://github.com/dengziming/kafka/tree/KAFKA-13959-2
Hey @dengziming, I took at look at the commits in this tree. Is this the
only commit
10 matches
Mail list logo
