Wonderful job tracking this down @lxbsz ! Do you have a test case for this? I realize it's racy but you could use a config to help trigger the race with barriers (similar to what we do with mds_inject_migrator_session_race)?

Not yet and I will add it.

 1988 void Client::encode_dentry_release(Dentry *dn, MetaRequest *req,
 1989                            mds_rank_t mds, int drop, int unless)
 1990 {
 1991   ldout(cct, 20) << __func__ << " enter(dn:"
 1992            << *dn << ")" << dendl;
 1993   int released = 0;
 1994   if (dn->dir)
 1995     released = encode_inode_release(dn->dir->parent_inode, req,
 1996                                     mds, drop, unless, 1);
 1997   if (released && dn->lease_mds == mds) {
 1998     ldout(cct, 25) << "preemptively releasing dn to mds" << dendl;
 1999     auto& rel = req->cap_releases.back();
 2000     rel.item.dname_len = dn->name.length();
 2001     rel.item.dname_seq = dn->lease_seq;
 2002     rel.dname = dn->name;
 2003     dn->lease_mds = -1;
 2004   }
 2005   ldout(cct, 25) << __func__ << " exit(dn:"
 2006            << *dn << ")" << dendl;
 2007 }

When doing the rename and in some cases, such as if the client_use_random_mds option is enabled, and if the replies of the rename operation are delayed for some reasons, such as delayed in the network route, then the Line#2003 above won't be executed and the _lookup(new dentry) will succeed. In this case both the new and old dentries will link the same inode at the same time.

This pull request can no longer be automatically merged: a rebase is needed and changes have to be manually resolved

Rebased it by resolving the conflict in commom/options.cc

jenkins test api

This pull request can no longer be automatically merged: a rebase is needed and changes have to be manually resolved

@lxbsz I'm seeing test failures relating to rename which makes me think if it has do with this change. Sample

• https://pulpito.ceph.com/vshankar-2022-09-13_09:03:58-fs-wip-vshankar-testing-20220909-120801-testing-default-smithi/7030450/
• https://pulpito.ceph.com/vshankar-2022-09-13_09:03:58-fs-wip-vshankar-testing-20220909-120801-testing-default-smithi/7030442/ (this one is stuck in a mv operation)

Sorry @vshankar , I missed this comment.

Currently I couldn't login the sepia node and couldn't checked the logs in detail, but from the teuthology.logs they looks like a similar issue with this.

I will check it more after I can login to the node.

jenkins test make check

@lxbsz Should I put this through QA? cc @vshankar

@lxbsz Should I put this through QA? cc @vshankar

Sure, please. Thanks.

This pull request has been automatically marked as stale because it has not had any activity for 60 days. It will be closed if no further activity occurs for another 30 days.
If you are a maintainer or core committer, please follow-up on this pull request to identify what steps should be taken by the author to move this proposed change forward.
If you are the author of this pull request, thank you for your proposed contribution. If you believe this change is still appropriate, please ensure that any feedback has been addressed and ask for a code review.

Updated the PR to fix one bug, which is when the rename request fails it will miss waking up the waiters:

11fb16d#diff-7a3052fe46aebfed0382c9d0bb9880ea1328add824e0b10c5d551ddfee282cd1R14464-R14469

@lxbsz Please let me know when this PR is ready. As per our chat one failure needs to be fixed.

@lxbsz Please let me know when this PR is ready. As per our chat one failure needs to be fixed.

Sure, I have gone through most failures, only the pjd ones are related to this PR, all the others are known issues. I will ping you after I finish that.

Thanks.

jenkins test docs

This PR is ready for merge, I'll merge it as soon as this CI job passes.