Hi Justin,
Thanks for creating the issue. I left a comment for you. Emailing you
here in case you didn't see it.
On Fri, May 30, 2025 at 7:02 PM Justin Chen
wrote:
>
> Here is the JIRA issue: https://issues.apache.org/jira/browse/KAFKA-19354
>
> Thank you,
> Justin C
>
> On Fri, May 30, 2025 at
Here is the JIRA issue: https://issues.apache.org/jira/browse/KAFKA-19354
Thank you,
Justin C
On Fri, May 30, 2025 at 11:47 AM José Armando García Sancio
wrote:
> Thanks for the bug report Justin. Looks like a bug to me. Please file
> a Jira with these details and share it in this thread.
>
> -
Thanks for the bug report Justin. Looks like a bug to me. Please file
a Jira with these details and share it in this thread.
--
-José
Hello,
Thank you for looking into this so quickly! Unfortunately, we didn't have
DEBUG logs enabled in observers nodes at the time as this was a production
environment.
However, I was able to reproduce the behaviour in a sandbox cluster with
the following setup (though the root cause may have diff
It's awesome you already reproduced the issue, Alyssa!
But @Justin, if possible, could you still share the logs and the quorum
state store file on the observer like Alyssa requested?
Thank you.
Luke
On Fri, May 30, 2025 at 9:40 AM Alyssa Huang
wrote:
> Wanted to correct my wording and resend t
Wanted to correct my wording and resend the original text of my email
because I got a bounce back -
> Thanks Justin,
>
> It's hard to say from the current details if it's simply a network issue
> (e.g. broker never receives the response with the leaderId), bug (broker
> does receive response with
Thanks Justin,
It's hard to say from the current details if it's simply a network issue
(e.g. broker never receives the response with the leaderId), bug (broker
does receive response with leaderId, never transitions to follower), or
something else. Could you potentially send over logs from the mis
To correct my original description:
We have observed that KRaft observers (process.roles=broker) that typically
send FETCH requests to the quorum Leader node can enter a state of
indefinitely **sending FETCH requests to a voter (follower) node**, which
we believe to be after a re-bootstrap due to
Hello,
In our Kafka 4.0 cluster (dynamic quorum, 5 controller nodes), we have
observed that KRaft observers (process.roles=broker) that typically send
FETCH requests to the quorum Leader node can enter a state of indefinitely
re-bootstraping to a voter (follower) node, likely after some sort of
re
