Hello Murilo,
I think what you've discovered is a processing spike: i.e. each poll call
itself returning up to 1000 records, and then processing them may take
longer (if, say, they hit the p99 percentile latency due to the processing
logic etc). After your config change, it is less sensitive to su
unsubscribe
On Thu, 14 Oct 2021 at 18:04, Murilo Tavares wrote:
> Hi Guozhang
> Thanks for your response. I'm on KafkaStreams 2.8.1.
> Since you asked, is KafkaStreams 3.0.0 compatible with a 2.4.1 broker?
>
> But I found the issue.
> TL;DR: I increased max.poll.interval.ms and decreased max.pol
Hi Guozhang
Thanks for your response. I'm on KafkaStreams 2.8.1.
Since you asked, is KafkaStreams 3.0.0 compatible with a 2.4.1 broker?
But I found the issue.
TL;DR: I increased max.poll.interval.ms and decreased max.poll.records which
fixed the problem.
I noticed the StreamThread summary logs
Pr
Hi Murilo, which version of Kafka Streams are you using? Could you try the
latest 3.0.0 release if it is not yet on that version?
On Wed, Oct 13, 2021 at 8:02 AM Murilo Tavares wrote:
> Hi
> I have a large, stateful, KafkaStreams application that is on a never
> ending rebalance loop.
> I can se
Hi
I have a large, stateful, KafkaStreams application that is on a never
ending rebalance loop.
I can see that Task restorations take a lng time (circa 30-45 min). And
after that I see this error.
This is followed by tasks being suspended, and the instance re-joining the
group and a new rebalan
