Afternoon all,
I've posted in the EMC forums but thought you lot might appreciate this also
Scenario:
- Novell SLES 11 SP1 Linux
- iSCSI transport
- Clariion AX4-5i
- 2x Database nodes in a HA cluster (SLES High Availability Extension / Pacemaker)
- Out of band STONITH via IBM IMM functional
- 2x Database roles which can move between servers to provide our HA environment
Key point 1: The HA cluster resources include the required SAN based filesystems, so only the filesystems required to run the database role will be mounted. ie: the filesystems are never mounted on more than one server at any given time. Should this happen, the HA monitor would detect and STONITH the offending node.
Key point 2: All SAN LUN's for both database rolls are visible to both servers all of the time, mounted or not.
Question Background:
Recently our production OCFS2 filesystems 'blew up' causing kernel panic's and subsequent server reboots. We have since reverted to ext3 as its proven, to do this safely we have opted to run our cluster one legged ... giving up our HA capability temporarily.
We have no issue in sticking with ext3, obviously we can never mount the filesystems on more than one node ... we have no requirement for this anyhow.
Question:
Is there any risk in having all SAN LUN's visible to both database servers all of the time but only ever mounted on one ? If there is I would be interested in hearing why ... to my knowledge the disk devices are probed only by the OS when not mounted etc.
Many thanks !!