The pr [6], which is a test-only change, should fix all the problems. It passes a 10x10 run [7].
Tsz-Wo [6] https://github.com/apache/ratis/pull/1287 [7] https://github.com/szetszwo/ratis/actions/runs/17886210393 On Sat, Sep 20, 2025 at 6:22 PM Tsz Wo Sze <[email protected]> wrote: > It turns out that RATIS-2261 is a test problem: > - It adds a majority of new servers causing the > problem described in RATIS-1930. > - It starts new servers using a group with all old and new peers -- if a > new server starts a leader election, the existing leader will ask it to > shut down since the new peer is not yet in the conf. > > Tsz-Wo > > > On Fri, Sep 19, 2025 at 9:01 AM Attila Doroszlai <[email protected]> > wrote: > >> > Maybe we should reconsider this pr[2] to make it more rubust. >> > [2] https://github.com/apache/ratis/pull/1275 >> >> I don't think it is caused by RATIS-2317, failure in >> TestInstallSnapshotNotificationWithGrpc can also be reproduced [3] >> with commits prior to that [4]. >> Also, for me it is still failing the repeated test [5] on master. >> >> -Attila >> >> [3] https://github.com/adoroszlai/ratis/actions/runs/17794561746 >> [4] >> https://github.com/apache/ratis/commit/7e22578d7af15cfcee9d1279025f644f420e4f36 >> [5] https://github.com/adoroszlai/ratis/actions/runs/17862737818 >> >
