Description of problem: When dir gfid entry is missing on few backend bricks for a directory, client heal is unable to re-create the gfid entries after doing stat from client. The automated test-case passes on downstream 3.4.4 but is failing on upstream gluster 6. Version-Release number of selected component (if applicable): Latest gluster 6 How reproducible: Always, Steps to Reproduce: 1. Create a 2X3 dist-replicated volume, and fuse mount it 2. Create a empty directory from mount point 3. Verify the gfid entry is present on all backend bricks for this dir 4. Delete gfid entry for 5 out of 6 backend bricks, brick{1..6} 5. Now trigger heal from mount pt. #ls -l #find . | xargs stat 6. Check backend bricks, the gfid entry should be healed for all the bricks. Actual results: At step 6, gfid entry is not created after client lookup. Expected results: Client lookup should trigger heal and gfid should be healed Additional info: There is also a latest fix per BZ#1661258, in which the dht delegates task to AFR when there is a missing gfid for all bricks in subvol, as per my understanding. The test-case is automated and can be found at https://review.gluster.org/c/glusto-tests/+/22480/ The test passes Downstream but fails upstream, the glusto logs for the failure can be found at https://ci.centos.org/job/gluster_glusto-patch-check/1277/artifact/glustomain.log
Created attachment 1553510 [details] glustomain_3_4_4.log
Marking it back to NEW, as the assignee is still bugs
This bug is moved to https://github.com/gluster/glusterfs/issues/916, and will be tracked there from now on. Visit GitHub issues URL for further details