Issue

• A kernel panic occurs after a gfs2 filesystem withdraws
• A node in our cluster was fenced right after it logged a GFS2 withdrawal error
• My cluster nodes using GFS2 are frequently panicking and dumping cores, showing GFS2 metadata inconsistency errors in the logs just before
• A kernel panic occurs after a cluster node was fenced by fence_scsi

[...]
[947302.618149] GFS2: fsid=tsmcluster:archlogfs.1: telling LM to unmount
[947302.618382] GFS2: fsid=tsmcluster:archlogfs.1: withdrawn
[...]
[947302.618519] GFS2: fsid=tsmcluster:archlogfs.1: Error -5 writing to log
[947302.618661] GFS2: lm_lock ret -22
[947302.618665]  G:  s:SH n:5/102bb f:lDpIqL t:UN d:UN/0 a:0 v:0 r:3 m:200
[947302.618697] ------------[ cut here ]------------
[947302.618698] kernel BUG at fs/gfs2/glock.c:546!
[947302.618699] invalid opcode: 0000 [#1] SMP
[...]

[....]
[  428.566409] GFS2: fsid=cluster1:gfs2fs.0: about to withdraw this file system
[  428.573202] GFS2: fsid=cluster1:gfs2fs.0: telling LM to unmount
[  428.574268] GFS2: fsid=cluster1:gfs2fs.0: withdrawn
[...]
[  428.574916] GFS2: lm_lock ret -22
[  428.575409]  G:  s:SH n:5/1f0421 f:lDpIqLo t:UN d:UN/0 a:0 v:0 r:3 m:200
[  428.575913] ------------[ cut here ]------------
[  428.576383] kernel BUG at fs/gfs2/glock.c:546!
[  428.576875] invalid opcode: 0000 [#1] SMP