Re: Kafka topic __consumer_offsets replication issue

[Luke Chen <sh...@gmail.com>]

Hi kiran,

I would check the log end offset of the in-sync partition first, and check
the lags of the offsets with the leader offset. (You can check by metric:
`kafka.log:type=Log,name=LogEndOffset,topic=xxx, partition=xx`)
Then, I would check if the follower is doing fetching without error by
checking the logs.
Basically, if the network is good, and storage system is good, and it's
doing fetching correctly, it should join in-sync soon.

Hope it works.

Thanks.
Luke

On Sun, Apr 30, 2023 at 10:58 AM kiran kumar <ki...@gmail.com> wrote:

> Hello All,
>
> Any insights on this issue?
>
> Thanks,
> Kiran
>
> On Sun, 23 Apr 2023 at 1:58 PM, kiran kumar <ki...@gmail.com>
> wrote:
>
> > Hello Kafka Folks,
> >
> > I have encountered an issue with Kafka 3.2 version.
> > Environment:
> > OS: RHEL 7.9
> > Java: 1.8
> > Kafka: 3.2
> >
> > There seems to one of the replica is stuck in __consumer_offsets topic.
> >
> > I have the replication factor set as 3 and its 5 node kafka cluster. I
> > could see that out of 2 replicas one replica is not getting in sync with
> > rest of the two replicas. I have trued to delete the partition folder and
> > performed the rolling restart of Kafka. I have also tried to reassign the
> > partition to another bode but that node as well in same state. I have
> tried
> > to expand the replicas to 5 partition reassignment tool but all the new
> > replicas are at the same state.
> > Even though, there two nodes which are good, the third node is unable to
> > completed sync and partitions to 5 nodes are also stuck at the same state
> > as the other faulty node.
> >
> > Is there any way to recover the Kafka partition replica and get it sync
> > with other two of the good nodes ?
> >
> > Thanks and regards
> > Kiran
> >
> > --
> > G.Kiran Kumar
> >
> --
> G.Kiran Kumar
>