Boot from San + Serviceguard Issue

[klemerS]

Hello.

We installed for a customer 2 systems, 11i V3
on Rx3600 + Boot from SAN + serviceguard 1.18

The issue is testing the cluster. all tests goes well , but disconnect the two fiber HBA from one system didnt cause the package to failover to the other system. We Create An EMS lv_summery resource , which was defined in our Cluster package , but still its seems that when the system disk is disconnected from the machine , nothing happen.

One way to solve the problem is to make mirror lvm of the system disk to a local sas disk.

But if someone have another IDEA how to get the system to reboot in case of such a case , i will appreciate

Regards

klemer Shalom.

[Duncan Edmonstone · Accepted solution]

Interesting issue - can you post the relevant configuration settings from your package configuration (how you define lv_summary as a resource dependency). I would usually set up a PVG for my VG in /etc/lvmpvg and use pv_summary rather than lv_summary

However I think there's a possibility you may be stuck here - normally in this situation I'd tell you to pull the fibres and then check the status of the EMS resource using the resls command. But of course in your scenario you won't be able to call the resls command as the system won't be able to page the executable in from disk... I wonder if Serviceguard suffers the same issue? If so, you may have no choice but to use those local disks...

You so know that when doing this kind of thing you have to have the NODE_FAIL_FAST_ENABLED set to true? As Serviceguard has no way of sensibly unmounting the hung filesystems, the only way to close down is to TOC the system.

If you have a support contract, you can look at the ITRC document UMCSGKBRC00012483 for a full example:

http://www4.itrc.hp.com/service/cki/docDisplay.do?docLocale=en&docId=emr_na-c01028614-3

HTH

Duncan

[klemerS]

n.