This is due to a race in lock mastery/purge. I have recently fixed this
problem but haven't yet submitted the patch to mainline. Please file a
Service request with Oracle to get a one-off fix.
On 02/25/2013 10:07 PM, Ramappa, Ravi (NSN - IN/Bangalore) wrote:
Hi,
In a 13 node cluster, the first four nodes went into kernel panic
state. The /var/log/messages contained messages as below,
Feb 25 22:02:46 prod152 kernel:
(o2net,9470,5):dlm_assert_master_handler:1817 ERROR: DIE! Mastery
assert from 4, but current owner is 2! (O000000000000000c36706200000000)
Feb 25 22:02:46 prod152 kernel: lockres:
O000000000000000c36706200000000, owner=2, state=0
Feb 25 22:02:46 prod152 kernel: last used: 0, refcnt: 3, on purge
list: no
Feb 25 22:02:46 prod152 kernel: on dirty list: no, on reco list: no,
migrating pending: no
Feb 25 22:02:46 prod152 kernel: inflight locks: 0, asts reserved: 0
Feb 25 22:02:46 prod152 kernel: refmap nodes: [ ], inflight=0
Feb 25 22:02:46 prod152 kernel: granted queue:
Feb 25 22:02:46 prod152 kernel: type=3, conv=-1, node=2,
cookie=2:222205, ref=2, ast=(empty=y,pend=n), bast=(empty=y,pend=n),
pending=(conv=n,lock=n,cancel=n,unlock=n)
Feb 25 22:02:46 prod152 kernel: converting queue:
Feb 25 22:02:46 prod152 kernel: blocked queue:
Feb 25 22:02:46 prod152 kernel: ----------- [cut here ] ---------
[please bite here ] ---------
Feb 25 22:02:46 prod152 kernel: Kernel BUG at
.../build/BUILD/ocfs2-1.4.10/fs/ocfs2/dlm/dlmmaster.c:1819
Feb 25 22:02:46 prod152 kernel: invalid opcode: 0000 [1] SMP
Feb 25 22:02:46 prod152 kernel: last sysfs file:
/block/cciss!c0d0/cciss!c0d0p1/stat
Feb 25 22:02:46 prod152 kernel: CPU 5
Feb 26 09:50:27 prod152 syslogd 1.4.1: restart.
The OCFS2 rpm versions used are as below,
[root@prod152 ~]# uname -r
2.6.18-308.1.1.el5
[root@prod152 ~]# rpm -qa| grep ocfs
ocfs2-2.6.18-308.1.1.el5xen-1.4.10-1.el5
ocfs2-tools-devel-1.6.3-2.el5
ocfs2-2.6.18-308.1.1.el5-1.4.10-1.el5
ocfs2-tools-debuginfo-1.6.3-2.el5
ocfs2-tools-1.6.3-2.el5
ocfs2console-1.6.3-2.el5
ocfs2-2.6.18-308.1.1.el5debug-1.4.10-1.el5
root@prod152 ~]# cat /etc/ocfs2/cluster.conf
node:
ip_port = 7777
ip_address = 10.10.10.150
number = 0
name = prod150
cluster = ocfs2
node:
ip_port = 7777
ip_address = 10.10.10.151
number = 1
name = prod151
cluster = ocfs2
node:
ip_port = 7777
ip_address = 10.10.10.152
number = 2
name = prod152
cluster = ocfs2
node:
ip_port = 7777
ip_address = 10.10.10.153
number = 3
name = prod153
cluster = ocfs2
node:
ip_port = 7777
ip_address = 10.10.10.106
number = 4
name = prod106
cluster = ocfs2
node:
ip_port = 7777
ip_address = 10.10.10.107
number = 5
name = prod107
cluster = ocfs2
node:
ip_port = 7777
ip_address = 10.10.10.155
number = 6
name = prod155
cluster = ocfs2
node:
ip_port = 7777
ip_address = 10.10.10.156
number = 7
name = prod156
cluster = ocfs2
node:
ip_port = 7777
ip_address = 10.10.10.157
number = 8
name = prod157
cluster = ocfs2
node:
ip_port = 7777
ip_address = 10.10.10.158
number = 9
name = prod158
cluster = ocfs2
node:
ip_port = 7777
ip_address = 10.10.10.51
number = 10
name = prod51
cluster = ocfs2
node:
ip_port = 7777
ip_address = 10.10.10.52
number = 11
name = prod52
cluster = ocfs2
node:
ip_port = 7777
ip_address = 10.10.10.154
number = 12
name = prod154
cluster = ocfs2
cluster:
node_count =13
name = ocfs2
[root@prod152 ~]#
Is this a known issue ? Any issues in the configuration ?
Thanks,
Ravi
_______________________________________________
Ocfs2-users mailing list
Ocfs2-users@oss.oracle.com
https://oss.oracle.com/mailman/listinfo/ocfs2-users
_______________________________________________
Ocfs2-users mailing list
Ocfs2-users@oss.oracle.com
https://oss.oracle.com/mailman/listinfo/ocfs2-users
