Description of problem: I have simple cluster: one storage exports GNBD and two nodes imports GNBD with LVM on top. Independently on mount options quota=on or quota=off I have trouble like this with suddenly freezes in any time after: Feb 27 07:16:46 node1 kernel: GFS2: fsid=cluster1:home.0: quotad: (2) error=-5 Feb 27 08:16:47 node1 kernel: GFS2: fsid=cluster1:home.0: warning: assertion "!gfs2_is_stuffed(ip)" failed Feb 27 08:16:47 node1 kernel: GFS2: fsid=cluster1:home.0: function = gfs2_block_map, file = fs/gfs2/bmap.c, line = 475 Feb 27 08:16:47 node1 kernel: [<f8b7a974>] gfs2_assert_warn_i+0x8d/0x117 [gfs2] Feb 27 08:16:47 node1 kernel: [<f8b5cfa9>] gfs2_block_map+0x77/0x3ab [gfs2] Feb 27 08:16:47 node1 kernel: [<f8b6a78d>] gfs2_log_reserve+0x106/0x117 [gfs2] Feb 27 08:16:47 node1 kernel: [<c046371c>] find_lock_page+0x1a/0x90 Feb 27 08:16:47 node1 kernel: [<f8b74002>] do_sync+0x340/0x58b [gfs2] Feb 27 08:16:47 node1 kernel: [<f8b73dcc>] do_sync+0x10a/0x58b [gfs2] Feb 27 08:16:47 node1 kernel: [<f8b74b13>] gfs2_quota_sync+0x1fe/0x281 [gfs2] Feb 27 08:16:47 node1 kernel: [<c043575e>] process_timeout+0x0/0x5 Feb 27 08:16:47 node1 kernel: [<f8b5e441>] gfs2_quotad+0xb1/0x140 [gfs2] Feb 27 08:16:47 node1 kernel: [<f8b5e390>] gfs2_quotad+0x0/0x140 [gfs2] Feb 27 08:16:47 node1 kernel: [<c043e32a>] kthread+0x38/0x5e Feb 27 08:16:47 node1 kernel: [<c043e2f2>] kthread+0x0/0x5e Feb 27 08:16:47 node1 kernel: [<c0405dbb>] kernel_thread_helper+0x7/0x10 Feb 27 08:16:47 node1 kernel: ======================= kernel is kernel-PAE-2.6.23.15-137.fc8.i686.rpm
That was fixed (incorrectly) a while back and a correct fix is now underway. You might find that the rawhide kernel will work with this, but it might be better to wait for the final fix which is expected shortly. I would have expected that turning off quotas would fix this, so the fact that you see this with quota=off is a bit strange.
The fix for this is upstream. Please reopen the bug if you continue to see this problem.