Oh also the other update, this happened with a cluster that is on 3.8.1. I still suspect a client bug rather than broker bug though.
On Wed, Mar 19, 2025 at 9:01 AM Donny Nadolny <do...@stripe.com> wrote: > Here's the jira: https://issues.apache.org/jira/browse/KAFKA-19012 > > I have a bit of new information as well: > - to your comment Alex, in our case the original message is not published > to the intended topic as it happened for you. Instead, the message only > ends up on the wrong topic > - Ismael: the client had permission to publish to any topic. It first > started happening around 4 months ago, because of the low frequency it's > tough to point to what changed that it's now triggered. The biggest recent > changes (upgrading broker version, upgrading kafka client version) all > happened > 6 months before it occurred. We did have other smaller changes > of varying closeness in time, but none of them seem to be the culprit (code > changes that increased the load average on the client machine -> we've > scaled out to lower the load and had no chance in frequency, various code > changes we temporarily reverted or deactivated but still saw occurrences, > etc). > > Thanks, > Donny > > > On Mon, Feb 10, 2025 at 10:25 AM Dhruvil Shah <dhru...@confluent.io.invalid> > wrote: > >> Could you let us know when the jira is filed? Please also include the >> producer and broker configs. >> >> On Sun, Feb 9, 2025 at 11:02 PM Oleksandr Shulgin < >> oleksandr.shul...@zalando.de> wrote: >> >> > On Sun, Feb 9, 2025 at 3:24 AM Ismael Juma <m...@ismaeljuma.com> wrote: >> > >> > > One more thing: when this happens, is the client authorized to write >> to >> > > both topicA and topicB? >> > > >> > >> > We, for one, do not use per-topic authorization, as it is enforced one >> > layer up. So the answer is "yes" in our case. >> > >> > -- >> > Alex >> > >> >
