Hi All, Running Ubuntu 16.04, with version JEWEL ceph version 10.2.2 (45107e21c568dd033c2f0a3107dec8f0b0e58374) In our environment we are running cephfs and our clients are connecting through ceph-fuse. Since I have upgraded from Hammer to Jewel I was haunted by the ceph-fuse segfaults, which were resolved by using the patch from https://github.com/ceph/ceph/pull/10027 But lately I'm often getting the error: "mds0: Client XXX:XXX failing to respond to capability release" These clients are always (patched) ceph-fuse clients and are generating the following MDS log error: 2016-09-14 08:06:36.953168 7f07df309700 0 log_channel(cluster) log [WRN] : client.196596070 isn't responding to mclientcaps(revoke), ino 10001d28611 pending pAsLsXsFscr issued pAsLsXsFsxcrwb, sent 30723.302011 seconds ago 2016-09-14 08:29:42.262518 7f07df309700 0 log_channel(cluster) log [WRN] : client.196596070 isn't responding to mclientcaps(revoke), ino 10001d28613 pending pAsxLsXsxFcb issued pAsxLsXsxFcwb, sent 62.097011 seconds ago 2016-09-14 08:30:42.263593 7f07df309700 0 log_channel(cluster) log [WRN] : client.196596070 isn't responding to mclientcaps(revoke), ino 10001d28613 pending pAsxLsXsxFcb issued pAsxLsXsxFcwb, sent 122.098144 seconds ago 2016-09-14 08:32:42.283509 7f07df309700 0 log_channel(cluster) log [WRN] : client.196596070 isn't responding to mclientcaps(revoke), ino 10001d28613 pending pAsxLsXsxFcb issued pAsxLsXsxFcwb, sent 242.118043 seconds ago 2016-09-14 09:37:32.347160 7f07df309700 0 log_channel(cluster) log [INF] : closing stale session client.196596070 10.5.5.83:0/4117427533 after 304.829698 It seems that these clients also trigger blocked requests. Everything is running Jewel code and I have no old clients anymore. My only solution is to restart the client (or forcebly remount). _______________________________________________ ceph-users mailing list ceph-users@xxxxxxxxxxxxxx http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com
 |