Thanks for the comment. Yes, it's indeed a downside that this will increase the size of the metadata response, I was thinking if it's worth providing this information to the Producer conditionally, e.g. add a config to Producer to allow people choose whether to include this information in the MetadataResponse or not. What do you think?
Br, Arvin Colin McCabe <cmcc...@apache.org> 于2020年7月7日周二 上午7:40写道： > Hi Arvin, > > Thanks for the KIP. > > Unfortunately, I don't think this makes sense since it would increase the > amount of data we send back in the metadata response, which is pretty bad > for scalability. In general we probably want to avoid the case where we > don't have the appropriate number of in-sync replicas, not try to optimize > for it. > > best, > Colin > > On Mon, Jul 6, 2020, at 10:38, Arvin Zheng wrote: > > Hi everyone, > > > > I would like to start the discussion for KIP-637 > > > https://cwiki.apache.org/confluence/display/KAFKA/KIP-637%3A+Include+min.insync.replicas+in+MetadataResponse+to+make+Producer+smarter+in+partitioning+events > > > > Looking forward to your feedback. > > > > Thanks, > > Arvin > > >
