Hi Chia-Ping, I agree that we have v3.9.3 for critical issues on ZK to KRaft migration. But like Mickael said, we should make it clear that we will have v3.9.4 or not? About this:
> So for 3.9.4, I think critical migration bugs are a good reason to make another release. I agree. But the "critical migration bug" is a vague definition to me. I think we can define it as: 1. The migration bug has no workaround. KAFKA-20488 <https://issues.apache.org/jira/browse/KAFKA-20488> is a good example that we document only because it has the workaround. 2. It will block the migration process, or cause the inconsistent metadata after migration. What do you think? Thanks, Luke On Wed, Nov 4, 2026 at 9:47 PM Chia-Ping Tsai <[email protected]> wrote: > Hi Mickael, > > Thanks for raising this. The most important issue to me is the migration > fixes, since 3.9 is the baseline for starting the migration. The CVE fixes > are more of an addendum. > > So for 3.9.4, I think critical migration bugs are a good reason to make > another release. Since 4.x no longer supports ZK, 3.9 is the only path for > ZK users to move to KRaft. If a bug blocks users from migrating, I believe > it is our responsibility as a community to fix it. CVE fixes can be > included when we have such a release, but they alone should not trigger a > new 3.9 release. > > Best, > Chia-Ping > > On Mon, Nov 02, 2026 11:43 PM, Mickael Maison <[email protected]> > wrote: > > > Hi, > > > > I'm not opposing making another 3.9 bugfix release but I think we > > should clarify our position regarding 3.9. > > 3.9.0 was released 2 years ago and it's not marked as supported anymore. > > > > I understand there's been a lot of CVEs published recently. However > > considering the current trends it's likely there will be more, so will > > we consider making a 3.9.4 release? > > The reason I'm raising this is to set clear expectations for our > > users, and to balance our workload as maintaining extra releases has a > > cost. > > > > Thanks, > > Mickael > > > > > > > > On Mon, Nov 2, 2026 at 2:32 PM Gaurav Narula <[email protected]> wrote: > > > > > > Thanks for taking the initiative Chia-Ping! > > > > > > I had refrained from commenting until now, having contributed to some > of > > the aforementioned fixes, thinking my position would have had some bias > but > > after the recent CVE announcements I strongly believe we should have a > > 3.9.3. > > > > > > Looking forward to the release plan! > > > > > > Regards, > > > Gaurav > > > > > > > On 2 Nov 2026, at 12:58, Chia-Ping Tsai via dev < > [email protected]> > > wrote: > > > > > > > > Hi all, > > > > > > > > We are going to backport the CVE fixes to 3.9 and I believe that is a > > good reason to make a 3.9.3 release. > > > > > > > > Hence, I will start the release plan for 3.9.3 and open another > thread > > once the plan is ready. > > > > > > > > Best, > > > > Chia-Ping > > > > > > > > On 2026-10-25 13:19 UTC Chia-Ping Tsai wrote: > > > >> Hi all, > > > >> > > > >> There are some critical bugs related to ZK migration, so I’d like to > > > >> open a discussion for the 3.9.3 release. > > > >> > > > >> KAFKA-20960: Block ZK-to-KRaft migration finalization until brokers > > > >> re-register (merged) > > > >> https://github.com/apache/kafka/commit/ea11d59ecda2ec57be0f1 > > ef83a554ef5d3b82770 > > > >> > > > >> KAFKA-21142: Fix partition creation when handling snapshot delta > > > >> during migration (merged) > > > >> https://github.com/apache/kafka/commit/daebfbd9b37c913937d5a > > dfc2e493fe825e3da3a > > > >> > > > >> KAFKA-21062: Handle BADVERSION on /migration writes (under review) > > > >> https://github.com/apache/kafka/pull/23639 > > > >> > > > >> My sense is that ZK migration is the only official way to encourage > > > >> users to get into KRaft, so keeping it working is a strong reason to > > > >> do a release. > > > >> > > > >> I haven’t finished reviewing the last one yet, but it would be good > to > > > >> make sure the release will happen. Otherwise, my tokens will be > > > >> totally wasted :) > > > >> > > > >> Any thoughts? > > > >> > > > > > >
