Thanks for the feedback Bruno. I updated the KIP (I thought the info was
there, but it seems it was too implicit).
1) Each time a client TimeoutException happens, we would log a WARN
message. If `task.timeout.ms` expires, we would rethrow the last client
`TimeoutException` to stop processing.
Thanks for the update, Matthias!
Other than Bruno’s good points, this proposal looks good to me.
Thanks,
John
On Thu, May 14, 2020, at 07:17, Bruno Cadonna wrote:
> Hi Matthias,
>
> Thank you for the KIP. I like your KIP.
>
> Here my feedback:
>
> 1. The KIP is not clear about what should
Hi Matthias,
Thank you for the KIP. I like your KIP.
Here my feedback:
1. The KIP is not clear about what should happen when task.timeout.ms
expires. To facilitate the mapping from the error users might
encounter due to timeouts to this KIP, it would be good to state the
error that will be
John, Guozhang,
thanks a lot for your feedback. I updated the KIP on a slightly
different angle: instead of using retries, we should switch to a timeout
based approach. I also extended the KIP to deprecate producer/admin
`retries` config.
I like the proposal to skip a task if a client
Hello John,
I'll make note that you owe me a beer now :)
I think I'm leaning towards your approach as well based on my observations
on previously reported timeout exceptions in the past. I once left some
thoughts on Matthias' PR here
Hi Matthias,
Thanks for the proposal! I think this will be a wonderful improvement
to Streams. In particular, thanks for the motivation. It would indeed
be nice not to have to set long timeout configs and block individual
client requests in order to cope with transient slow responses.
I'm very
-BEGIN PGP SIGNED MESSAGE-
Hash: SHA512
Hi,
I would like to propose KIP-572 to make Kafka Streams more robust with
regard to timeout exception handling.
https://cwiki.apache.org/confluence/display/KAFKA/KIP-572%3A+Improve+tim
eouts+and+retries+in+Kafka+Streams
Note, that there is a