Hey Sean, You can do rolling restart, then can do a describe of the topic, to check if the broker is back. Then continue with the next one.
Best, Xin On 13.04.18, 21:09, "McNealy, Sean" <sean.mcne...@ncr.com> wrote: I have a cluster of 3 kafka brokers running 0.10.2 that have been running for a while with the property “log.cleaner.enable=false” and I would like to turn compaction on. Is it safe to do this using a rolling restart? Will the cluster stay healthy with some brokers running with the compaction property turned on and some still off? Will some partitions start compacting and some not yet? Can partitions stay in sync? Or does this property require every broker to be stopped simultaneously, then make the change, and then started. Thanks, Sean