How long does this last? Maybe there are some async commit-offset frames left in the send buffer to that node? Can you reliably reproduce this?
On Wed, Mar 23, 2016 at 6:29 AM, Saurabh Daftary <saurabh.daft...@gmail.com> wrote: > I am running broker 0.9.0 with consumer 0.9.0.1. > > Seeing an issue wherein after a consumer rebalance (when I add a new > consumer) my old consumer still keeps committing offsets to partition it no > longer holds after a rebalance. > > The rebalance seem to work ok - On run ConsumerGroupCommand I see that the > partitions are redistributed after I add a new consumer. The respective > consumers do their fetches only on their assigned partitions but the old > consumer still commits the same last offset it had when it was holding on > to that partition At the same time the added consumer commits the correct > offsets. > > Any inputs on what could be going on? > -- *Christian Posta* twitter: @christianposta http://www.christianposta.com/blog http://fabric8.io