Denis Otkidach
Denis Otkidach
@tvoinarovskyi Please review
@kbhatiya999 Could you please describe what problem do you solve? Is it possible to reproduce it?
@kbhatiya999, thank you for the reminder about this problem. I've done some research and it looks like the main point of the PR is correct, i.e. we have to send...
> * drop the supervisord approach in docker > * create a docker-compose file that starts zookeeper, a kerberos server, and 3 brokers I completely agree with moving to docker...
I think some temporary docker-compose.yml to reproduce the problem and check that PR fixes it (without automating and incorporating it in CI) would be enough for now.
I see the Jira issue ([KAFKA-9705 - Zookeeper mutation protocols should be redirected to Controller only](https://issues.apache.org/jira/browse/KAFKA-9705)) is not resolved yet.
> I am not sure then how far the goal of aiokafka client to be compatible with different servers I don't mind adding a recent version of RedPanda to CI,...
@Symas1 Your approach may not work if you send messages quickly enough, as aiokafka combines multiple messages into batches and raising this setting may lead to huge requests being rejected...
~Looks like a bug to me.~
> The same parameters(consumer_timeout_ms), why the effect is not the same I believe Vincent meant to use `timeout_ms` parameter of `getmany`, not `consumer_timeout_ms `.