Hello Boyang, Your observation is valid, in fact after KIP-134 is merged there are some discussions about on the voting thread as well, that whether we should better make it a client config than a global broker-level config.
There are some general ideas grooming to further improve consumer rebalance protocols, including this, and motivations of KAFKA-7126 / KAFKA-7018 / as well. Please stay tuned. Guozhang On Sun, Jul 1, 2018 at 6:45 PM, Boyang Chen <bche...@outlook.com> wrote: > Hey there, > > > I have a question regarding KIP-134<https://cwiki.apache. > org/confluence/display/KAFKA/KIP-134%3A+Delay+initial+ > consumer+group+rebalance>. Basically this config `group.initial.rebalance. > delay.ms` is set on broker side to delay the initial rebalance of > consumer groups. For certain consumers, we would like to have customized > group initial rebalance delay, since for some testing classes the time to > bring up hosts (just 1) is minimal. We could code this delay in the first > member join request, so that broker could start rebalancing faster than > cluster level config. Does that make sense? > > > Best, > > Boyang > > > > -- -- Guozhang
