Failover with bad LUN

[Stuart Powell]

In our Serviceguard development 2 node cluster we experienced a disk array failure that took out a LUN. The LUN has no data in it, so we are not concerned with data integrity. However the LUN is included in a serviceguard volume group. There are 3 LUN, 6 disks, 3 disks, and 3 disks in each LUN. The bad LUN has 3 disks, so the vg has quorum.
To fix the LUN we need to bring the application down, but it is a production applicaiton, so we are waiting for the right time. What should happen if the package failes on its current node before we repair the LUN?

[Sanjay_6 · Accepted solution]

Hi Stuart,

Since the quorum is there, even if the package fails to the adoptive node, the vg should get ativated and the used filesystem would get mounted. Was there any mount configured on the failed lun. If so the package may not start on the adoptive node, otherwise everyhing should be okay. What about the disk redundancy in the array ?.

Hope this helps.

Regds

[Stephen Doud]

Hey Stuart,

I agree with Sanjay on both counts - even if the package fails over, the VG still meets quorum (assuming >50% of LUNs are available even after the package failover) so the package should start.
As Sanjay states, if the LUN is normally mounted, it may fail the check_and_mount function in the control script and immediately halt again.

-s.

[Stuart Powell]

There was no data on the LUN, so we don't have to worry about package access.
Thanks for the input.