On Thu, Jul 14, 2022 at 4:55 PM Jason Gustafson <ja...@confluent.io.invalid> wrote: > > Hey Jose, > > Thanks for volunteering to manage the release! KIP-833 is currently slotted > for 3.3. We've been getting some help from Jack Vanlighty to validate the > raft implementation in TLA+ and with frameworks like Jepsen. The > specification is written here if anyone is interested: > https://github.com/Vanlightly/raft-tlaplus/blob/main/specifications/pull-raft/KRaft.tla. > The main gap that this work uncovered in our implementation is documented > here: https://issues.apache.org/jira/browse/KAFKA-14077. I do believe that > KIP-833 depends on fixing this issue, so I wanted to see how you feel about > giving us a little more time to address it?
Thanks Jason. Ismael, Jason, Colin and I discussed this offline. We don't think this should be a blocker for 3.3.0. "KIP-853: KRaft Voter Changes" is my proposal to fix this. If the KIP is approved, we should be able to include the fix in 3.4.0. I went ahead and replaced the fix version with 3.4.0. For 3.3.0 we planned to improve the kafka-metadata-quorum tool so that it can tell the Apache Kafka administrator when it is safe to bring back a controller with a failed disk. -- -José
