statd_callout_helper processes client state for all public IP addresses, not only the ones taken during the current takeover run. This means some of the clients will not reclaim locks and the processed client state is deleted. Therefore, those client will not be notified when relevant public IPs are later failed over.
This bug was referenced in samba master: 85afee0a83dd2f70b90cff4c1e21b865640261fb e4914e6a4f1cb77eebf86c5ab3f241c2a9e5bd05
Created attachment 18853 [details] Patch for v4-24-test, v4-23-test, v4-22-test Commits cherry-pick cleanly into v4-24-test. Resulting patch also applies cleanly to v4-23-test and v4-22-test. Note that this must be applied before the patch attached to BZ15938.
Also, affected unit tests pass on 4.22.
Re-assigning for inclusion in 4.22, 4.23 and 4.24.
This bug was referenced in samba v4-24-test: 3ab52bf550ebe354af9c3aa18a0167e070caa3d4 e1bf06d0d5ebdf927b8a738025c608c8e39452a1
This bug was referenced in samba v4-24-stable (Release samba-4.24.0rc3): 3ab52bf550ebe354af9c3aa18a0167e070caa3d4 e1bf06d0d5ebdf927b8a738025c608c8e39452a1
Hi Björn, Did you mean to close this? It isn't yet in 4.23 or 4.22. :-(
Hi Martin, Günther asked a similar question yesterday in bugzilla :). I close the bug as soon as I've pushed the fixes to the required release branches. If autobuild fails due to flapping tests (which happend more often that it succeeds unfortunately), it will be delayed until the next push. I can't do the book keeping for all the failing flapping tests and keep track of the assigned bugs and when all the builds finally succeeded. I file bugs for the flapping tests though, so that they hopefully get fixed and the whole process with autobuild becomes less painful.
This bug was referenced in samba v4-23-test: 2a1a4ca22e93fe40788871df3de3b05247755246 7cf0f098b35e9f0d11fc4b8646409821067d01d5
This bug was referenced in samba v4-23-stable (Release samba-4.23.6): 2a1a4ca22e93fe40788871df3de3b05247755246 7cf0f098b35e9f0d11fc4b8646409821067d01d5