On Sat, Feb 13, 2016 at 8:14 AM, Blade Doyle <blade.doyle@xxxxxxxxx> wrote: > Greg, Thats very useful info. I had not queried the admin sockets before > today, so I am learning new things! > > on the x86_64: mds, mon, and osd, and rbd + cephfs client > ceph version 0.94.5 (9764da52395923e0b32908d83a9f7304401fee43) > > On the arm7 nodes: mon, osd, and rbd + cephfs clients > ceph version 0.94.3 (95cefea9fd9ab740263bf8bb4796fd864d9afe2b) > > Yes, all reported ops are on the same inode. > > I power cycled the entire cluster last night, including mds and clients. > That *did* clear up the read locks and I was able to access the files again. > > However, the stuck read locks have returned ;( So now I have the same > issue again. > > It seems likely that the "plex" media player is the client that holds the > initial hanging read lock. It could be something specific that app is > doing, or could just be a coincidence. > > It would be reasonably easy to update ceph version on the x86_64 (mds > server, a mon, and the client running plex). I'll work on that if you think > it could solve my issues? It might. If it doesn't I'd be interested in whatever logs and debugging information you can share with upstream to try and identify what's going on. :) -Greg > > Thanks again! > Blade. > > _______________________________________________ > ceph-users mailing list > ceph-users@xxxxxxxxxxxxxx > http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com > _______________________________________________ ceph-users mailing list ceph-users@xxxxxxxxxxxxxx http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com
 |