Hi, I would appreciate it if some one could answer this question. We have three nodes in a GFS cluster: NODE1: GFS-6.0.2-25 NODE2: GFS-6.0.2.36-1 NODE3: GFS-6.0.2.36-1 NODE2 does not always invoke "lock_gulmd: shutdown" during every reboot. When it does not, NODE1 (master) will think NODE2 misses heartbeats and then fence NODE2 because NODE1 does not know NODE2 is rebooting. Please see the kernel log on NODE2 below. The first part of the log shows that NODE2 does not invoke "lock_gulmd: shutdown" during the first reboot (then it got fenced by NODE1). The second part shows that NODE2 invokes "lock_gulmd: shutdown" properly during the second reboot (then NODE2 can re-join the cluster automatically after bootup). First part: Feb 23 16:59:12 NODE2 crond: crond shutdown succeeded Feb 23 16:59:12 NODE2 lock_gulmd_core[6633]: "GFS Kernel Interface" is logged out. fd:9 Feb 23 16:59:12 NODE2 gfs: Unmounting GFS filesystems: succeeded Feb 23 16:59:14 NODE2 ccsd[6614]: Stopping ccsd, SIGTERM received. Feb 23 16:59:14 NODE2 ccsd: Stopping ccsd: Feb 23 16:59:16 NODE2 ccsd: shutdown succeeded Feb 23 16:59:16 NODE2 ccsd: _[60G Feb 23 16:59:16 NODE2 ccsd: Feb 23 16:59:16 NODE2 rc: Stopping ccsd: succeeded Feb 23 16:59:16 NODE2 dd: 1+0 records in Feb 23 16:59:16 NODE2 dd: 1+0 records out Feb 23 16:59:16 NODE2 random: Saving random seed: succeeded Feb 23 16:59:16 NODE2 mdmonitor: mdadm shutdown succeeded Feb 23 16:59:16 NODE2 pool: Stopping pool pool_ccs: Feb 23 16:59:16 NODE2 pool: shutdown succeeded Feb 23 16:59:16 NODE2 pool: shutdown succeeded Feb 23 16:59:16 NODE2 pool: Feb 23 16:59:16 NODE2 pool: Feb 23 16:59:16 NODE2 rc: Stopping pool: succeeded Second part: Feb 23 17:07:40 NODE2 crond: crond shutdown succeeded Feb 23 17:07:41 NODE2 lock_gulmd_core[5189]: "GFS Kernel Interface" is logged out. fd:9 Feb 23 17:07:41 NODE2 gfs: Unmounting GFS filesystems: succeeded Feb 23 17:07:43 NODE2 lock_gulmd: Checking for Gulm Services... Feb 23 17:07:43 NODE2 lock_gulmd: Stopping lock_gulmd: Feb 23 17:07:44 NODE2 lock_gulmd: shutdown succeeded Feb 23 17:07:44 NODE2 lock_gulmd: _[60G Feb 23 17:07:44 NODE2 lock_gulmd: Feb 23 17:07:44 NODE2 rc: Stopping lock_gulmd: succeeded Feb 23 17:07:44 NODE2 ccsd[5149]: Stopping ccsd, SIGTERM received. Feb 23 17:07:44 NODE2 ccsd: Stopping ccsd: Feb 23 17:07:46 NODE2 ccsd: shutdown succeeded Feb 23 17:07:46 NODE2 ccsd: _[60G Feb 23 17:07:46 NODE2 ccsd: Feb 23 17:07:46 NODE2 rc: Stopping ccsd: succeeded Feb 23 17:07:46 NODE2 dd: 1+0 records in Feb 23 17:07:46 NODE2 dd: 1+0 records out Feb 23 17:07:46 NODE2 random: Saving random seed: succeeded Feb 23 17:07:46 NODE2 mdmonitor: mdadm shutdown succeeded Feb 23 17:07:46 NODE2 pool: Stopping pool pool_ccs: Feb 23 17:07:46 NODE2 pool: shutdown succeeded Feb 23 17:07:46 NODE2 pool: shutdown succeeded Feb 23 17:07:46 NODE2 pool: Feb 23 17:07:46 NODE2 pool: Feb 23 17:07:46 NODE2 rc: Stopping pool: succeeded The question is what could be the reason for NODE2 not to invoke "lock_gulmd: shutdown" during a reboot? Best regards, BJ -- ____________________________________________________________________________________ Be a better friend, newshound, and know-it-all with Yahoo! Mobile. Try it now. http://mobile.yahoo.com/;_ylt=Ahu06i62sR8HDtDypao8Wcj9tAcJ -- Linux-cluster mailing list Linux-cluster@xxxxxxxxxx https://www.redhat.com/mailman/listinfo/linux-cluster