You are viewing a plain text version of this content. The canonical link for it is here.

Posted to dev@pulsar.apache.org by ma...@gmail.com on 2023/03/09 03:44:19 UTC

[DISCUSS] Critical problem report - session notification thread deadlock

Hi, All

We found a critical problem that will cause the pulsar cluster to part “deaf” status. the broker can’t receive zookeeper session notification to revalidate namespace bundle ownership and leader election. That means it will cause one topic may have two owner brokers, and the leader election problem. etc

Blast radius:
Since this problem was introduced by https://github.com/apache/pulsar/pull/17401. The releases as follows may affect:

2.8.5
2.9.4
2.10.3
2.11.0

Workaround:

Restart broker

The fix is over here: https://github.com/apache/pulsar/pull/19754

We can avoid upgrading to the above version and wait for the latest progress. I will continue to push for a solution to this problem.

Please correct me if I got something wrong. thanks!


Best,
Mattison

Re: [DISCUSS] Critical problem report - session notification thread deadlock

Posted by PengHui Li <pe...@apache.org>.

Thanks, Qiang,

And for the upcoming patch releases

2.11.1,
2.10.4,
2.9.5

Please ship the fix to the release. Although it's not a fix for breaking
change.
But it fixed a critical issue from previous releases.

Thanks,
Penghui

On Thu, Mar 9, 2023 at 11:44 AM <ma...@gmail.com> wrote:

> Hi, All
>
> We found a critical problem that will cause the pulsar cluster to part
> “deaf” status. the broker can’t receive zookeeper session notification to
> revalidate namespace bundle ownership and leader election. That means it
> will cause one topic may have two owner brokers, and the leader election
> problem. etc
>
> Blast radius:
> Since this problem was introduced by
> https://github.com/apache/pulsar/pull/17401. The releases as follows may
> affect:
>
> 2.8.5
> 2.9.4
> 2.10.3
> 2.11.0
>
> Workaround:
>
> Restart broker
>
> The fix is over here: https://github.com/apache/pulsar/pull/19754
>
> We can avoid upgrading to the above version and wait for the latest
> progress. I will continue to push for a solution to this problem.
>
> Please correct me if I got something wrong. thanks!
>
>
> Best,
> Mattison
>