Description of problem: A blocking lock request (e.g. a conflicting lock request on a file region using flock(2)) will result in MDS reporting/logging the request as "slow request". Version-Release number of selected component (if applicable): 3.0 How reproducible: 100% Steps to Reproduce: Best tested by using flock(1) (as below) or following similar steps w/ a C program. 1. In one ceph kclient mount run: flock -x </path/to/file/in/cepfs/mount> sleep 3600 2. In another ceph kclient mount run: flock -x </path/to/file/in/cepfs/mount> sleep 3600 3. wait for some time (~30 odd seconds) 4. Check the MDS log for slow requests (should resemble something like below): 2018-06-25 03:27:30.305830 7f039ad85700 0 log_channel(cluster) log [WRN] : 1 slow requests, 1 included below; oldest blocked for > 32.606567 secs 2018-06-25 03:27:30.305895 7f039ad85700 0 log_channel(cluster) log [WRN] : slow request 32.606567 seconds old, received at 2018-06-25 03:26:57.699199: client_request(client.324120:4 setfilelockrule 2, type 2, owner 15445109620320908090, pid 11253, start 0, length 0, wait 1 #0x10000000001 2018-06-25 03:26:57.692840 caller_uid=1000, caller_gid=1000{}) currently acquired locks Additionally the blocked request can be checked by dumping the in-flight request as: # ceph daemon mds.<id> dump_ops_in_flight Actual results: Slow request reported by MDS: Expected results: The lock request should not be reported as slow request
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