Hey Jason, Thanks for the KIP. The KIP looks good overall.
I have one minor question. The KIP says that "If the producer sees either the FENCED_LEADER_EPOCH or the UNKNOWN_LEADER_EPOCH in the response from the broker, it will refresh metadata and retry". Given that UNKNOWN_LEADER_EPOCH is returned to producer iff producer's leaderEpoch > broker's leaderEpoch, it probably means the producer's metadata is newer than broker's leadership information. In this case, does producer needs to refresh metadata? Thanks, Dong On Tue, Aug 28, 2018 at 9:07 AM, Jason Gustafson <[email protected]> wrote: > Hey All, > > I didn't expect many comments here. This is mostly for consistency and to > improve debugging now that we have leader epoch in the metadata. I'll go > ahead and start a vote shortly. > > Thanks, > Jason > > On Sat, Aug 18, 2018 at 2:10 PM, Jason Gustafson <[email protected]> > wrote: > > > Hi All, > > > > I've added a short KIP to add leader epoch validation to the produce API. > > This is a follow-up to KIP-320, which added similar protection to the > > consumer APIs. Take a look and let me know what you think. > > > > https://cwiki.apache.org/confluence/display/KAFKA/KIP- > > 359%3A+Verify+leader+epoch+in+produce+requests > > > > Thanks, > > Jason > > >
