+1. I am in favor of declaring support for JDK 21. I would like to point out that daily tests are run on JDK 21 in TeamCity: while there are indeed more test failures than on JDK 17, Alexander Nikolaev is aware of the situation and has already fixed some of them.
On Wed, Sep 30, 2026 at 3:46 PM Nikolaev Aleksandr <[email protected]> wrote: > Hi, Ignite has been running on JDK 21 since 2.16. The compilation issues > were mostly in tests and Javadoc generation, not in the product. The tests > used APIs that are no longer available in JDK 21, so the fixes were about > updating the tests, not fixing broken functionality. As for IGNITE-29086, > it's in performance statistics, not the core. The release dates aren't > final yet, so there's still time to fix it before the release. If we don't > make it, it can go into a hotfix. We'll never be able to say everything has > been checked and there are no more bugs. Even the JDK itself gets regular > bugfix releases for every version. I don't see he point in delaying the > announcement further. We just need to respond to such issues faster, with > more frequent releases or hotfixes. > ________________________________ > От: Aleksandr Chesnokov <[email protected]> > Отправлено: 30 сентября 2026 г. 12:23 > Кому: [email protected] <[email protected]> > Тема: Re: Apache Ignite 2.19.0 RELEASE [Time, Scope, Manager] > > Hi all, > > +1 for the new release (non-binding). But I have some doubts about JDK 21 > support. > > Two months ago Ignite couldn't even be compiled on JDK 21 ( > https://issues.apache.org/jira/browse/IGNITE-28975). > > Also, a few days ago I found new JDK 21 bug related to performance > statistics and I'm still investigating it: > https://issues.apache.org/jira/browse/IGNITE-29086 > > So could you clarify why you think we are ready to declare JDK 21 support? > The fact that a number of JDK 21 fixes have already landed in master > doesn't mean we have found all the bugs. It looks like we haven't looked > for new ones enough yet. > > Best regards, > > Aleksandr Chesnokov >
