[Linux-cluster] cluster did not fail over when san mount was lost

[Neil Watson]

I arranged for a cluster's shared partition (/db2, ext3, attached via
HBA to a SAN) to be disconnected while the cluster was busy.  I was
expecting the cluster to fail over to the standby node.  This did not
happen.  From what I can tell the kernel of the active node reports the
file system error and then then remounts /db2 as read only.  The fs.sh
script seems to check that the /db2 mount exists but, does not check its
state.  Thus, fs.sh reports nothing wrong.  Log entries:

Aug 25 11:38:59 caesar kernel: EXT3-fs error (device sda1): ext3_journal_start_sb: Detected aborted journal
Aug 25 11:38:59 caesar kernel: Remounting filesystem read-only
Aug 25 11:39:00 caesar clurgmgrd: [5719]: <crit> Finding status of: /dev/sda1 /db2  
Aug 25 11:39:00 caesar clurgmgrd: [5719]: <crit> in isMounted with /dev/sda1 and /db2  
Aug 25 11:39:00 caesar clurgmgrd: [5719]: <crit> isMounted returned: 0  
Aug 25 11:39:00 caesar clurgmgrd: [5719]: <crit> Past isMounted  
Aug 25 11:39:00 caesar clurgmgrd: [5719]: <crit> running isAlive with: /db2  
Aug 25 11:39:00 caesar clurgmgrd: [5719]: <crit> isAlive returnd: 0  
Aug 25 11:39:20 caesar clurgmgrd: [5719]: <crit> Finding status of: /dev/sda1 /db2  
Aug 25 11:39:20 caesar clurgmgrd: [5719]: <crit> in isMounted with /dev/sda1 and /db2  
Aug 25 11:39:20 caesar clurgmgrd: [5719]: <crit> isMounted returned: 0  
Aug 25 11:39:20 caesar clurgmgrd: [5719]: <crit> Past isMounted  
Aug 25 11:39:20 caesar clurgmgrd: [5719]: <crit> running isAlive with: /db2  
Aug 25 11:39:20 caesar clurgmgrd: [5719]: <crit> isAlive returnd: 1  

Is this a bug or, have I missed something?

Configuration is 2 systems in active, standby mode.  Both are AS4 x86_64
and are currently up to date with their patches.

-- 
Neil Watson             | Gentoo Linux
System Administrator    | Uptime 29 days
http://watson-wilson.ca | 2.6.16.19 AMD Athlon(tm) MP 2000+ x 2