Hi Thomas, Thanks for the KIP. Overall, the KIP looks good to me.
I have only one question: The FindCoordinator API only supports resolving one group id at the time. If we want to get the offsets for say N groups, that means that we have to first issue N FindCoordinator requests, wait for the responses, group by coordinators, and then send a OffsetFetch request per coordinator. I wonder if we should also extend the FindCoordinator API to support resolving multiple groups as well. This would make the implementation in the admin client a bit easier and would ensure that we can handle multiple groups end-to-end. Have you thought about this? Best, David On Tue, Jan 26, 2021 at 10:13 AM Rajini Sivaram <rajinisiva...@gmail.com> wrote: > Hi Thomas, > > Thanks for the KIP, this is a useful addition for admin use cases. It may > be worth starting the voting thread soon if we want to get this into 2.8.0. > > Regards, > > Rajini > > > On Mon, Jan 25, 2021 at 1:52 PM Thomas Scott <t...@confluent.io> wrote: > > > Thanks Ismael, that's a lot better. I've updated the KIP with this > > behaviour instead. > > > > On Mon, Jan 25, 2021 at 11:42 AM Ismael Juma <ism...@juma.me.uk> wrote: > > > > > Thanks for the KIP, Thomas. One question below: > > > > > > Should an Admin client with this new functionality be used against an > old > > > > broker that cannot handle these requests then the methods will throw > > > > UnsupportedVersionException as per the usual pattern. > > > > > > > > > Did we consider automatically falling back to the single group id > request > > > if the more efficient one is not supported? > > > > > > Ismael > > > > > > On Mon, Jan 25, 2021 at 3:34 AM Thomas Scott <t...@confluent.io> wrote: > > > > > > > Hi, > > > > > > > > I'm starting this thread to discuss KIP-709 to extend OffsetFetch > > > requests > > > > to accept multiple group ids. Please check out the KIP here: > > > > > > > > > > > > > > https://cwiki.apache.org/confluence/pages/viewpage.action?pageId=173084258 > > > > > > > > Any comments much appreciated. > > > > > > > > thanks, > > > > > > > > Tom > > > > > > > > > >
