Description of problem: If clients get evicted in the middle of mds failover，there is a chance that ceph-mds crashes. How reproducible: Steps to Reproduce: 1. Setup 3 active mds, 1 ceph-fuse client, run metadata extensive workload (such as fsstss -p 16 ...) 2. Restart 2 mds. When one recovering mds enter rejoin state, kill the client by using 'ceph daemon mds.xxx session evict xxx' Actual results: ceph-mds may crash Expected results: ceph-mds does not crash Additional info:
zhyan-laptop:ceph zhyan$ git push rh-ceph ceph-3.0-rhel-patches Counting objects: 62, done. Delta compression using up to 4 threads. Compressing objects: 100% (57/57), done. Writing objects: 100% (62/62), 11.35 KiB | 726.00 KiB/s, done. Total 62 (delta 54), reused 5 (delta 5) remote: Resolving deltas: 100% (54/54) remote: Processing changes: refs: 1, done To ssh://code.engineering.redhat.com/ceph/.git 0b3592f04c..1e9171374c ceph-3.0-rhel-patches -> ceph-3.0-rhel-patches
MDS version: ceph version 12.2.4-45redhat1xenial followed the steps from desc,mds no longer crashes when the client is evicted. Moving to verified state
Since the problem described in this bug report should be resolved in a recent advisory, it has been closed with a resolution of ERRATA. For information on the advisory, and where to find the updated files, follow the link below. If the solution does not work for you, open a new bug report. https://access.redhat.com/errata/RHBA-2018:2375