Andreas
I just checked the log carefully, and this seems a different problem.
The reason of this failure is that MDT can not be start, which seems related with Liang's pdir patch. I will reassign this to Liang.
17:36:06:kmem_cache_create: duplicate cache dynlock_cache
17:36:07:Pid: 10192, comm: modprobe Not tainted 2.6.32-220.el6_lustre.x86_64 #1
17:36:08:Call Trace:
17:36:08: [<ffffffff81161da8>] ? kmem_cache_create+0x538/0x5a0
17:36:08: [<ffffffff81161cc6>] ? kmem_cache_create+0x456/0x5a0
17:36:09: [<ffffffffa06e4eb0>] ? init_once+0x0/0x90 [ldiskfs]
17:36:09: [<ffffffffa046c30d>] ? dynlock_cache_init+0x1f/0x42 [ldiskfs]
17:36:09: [<ffffffffa046c2af>] ? init_ldiskfs_fs+0x1d7/0x1e2 [ldiskfs]
17:36:10: [<ffffffffa046c0d8>] ? init_ldiskfs_fs+0x0/0x1e2 [ldiskfs]
17:36:12: [<ffffffff8100204c>] ? do_one_initcall+0x3c/0x1d0
17:36:12: [<ffffffff810af641>] ? sys_init_module+0xe1/0x250
17:36:12: [<ffffffff8100b0f2>] ? system_call_fastpath+0x16/0x1b
17:36:13:Not able to create dynlock cache
17:36:14:LustreError: 10167:0:(obd_mount.c:1445:server_kernel_mount()) premount /dev/mapper/lvm--MDS-P0:0x0 ldiskfs failed: -19 Is the ldiskfs module available?
17:36:14:LustreError: 10167:0:(obd_mount.c:1770:server_fill_super()) Unable to mount device /dev/mapper/lvm--MDS-P0: -19
17:36:15:LustreError: 10167:0:(obd_mount.c:2316:lustre_fill_super()) Unable to mount (-19)
17:36:15:Lustre: DEBUG MARKER: conf-sanity test_53b: @@@@@@ FAIL: MDT start failed
It seems this is a duplicate of
LU-1048, not by pdir ops patch. Sorry.