EMC PowerPath or DM-MPIO which is used in redhat HA-cluster

nasiruddin · 02-03-2012, 08:27 AM

we have a two node redhat HA cluster have some 5-6 LUNS on the SAN shared between the two.Both nodes have RHEL 5.5
Last week there was a kernel error in one of the node and failover was also not successful.
On review we come to know that both(multipath software) DM-MPIO and EMC PowerPath is running.
Can some body tell me that how to know which one is RH HA-cluster is using and which one should i stop so that this error does not occurs again.
I would also like to know how it can be stopped?

Please Help....its urgent

thanks in advance

grim76 · 02-03-2012, 09:06 AM

Which multipathing you are using depends on how you are referring and using the luns. Typically devices for powerpath are emcpower<a-z>. What devices you are using in your configuration should tell you which one you will need to keep.

FYI - Just to let you know putting urgent in the post won't get you help any faster. This is a volunteer based forum. In the event that you need assistance that quickly I would suggest that you contact EMC, and or Red Hat. That way you can use your support contracts to get priority support.

nasiruddin · 02-06-2012, 12:46 AM

Thanks for your reply grim76....
well you were correct that the devices are named as emcpower<a-z>.but we created volume groups,physical volume and then logical volumes on it and therefore used as

/dev/<volume-group>/<logical-volume> in our clustering environment

can you now tell me what should be my approach?

thanks once again...

p.s:- I know this is a volunteer based forum...pardon me if my urgent tag hurt you....

grim76 · 02-06-2012, 07:33 AM

What devices your PV, VG, and LV are using would be related to how LVM is configured. LVM by default scans all the available devices looking for LVM related items. The problem with that is typically it will find /dev/sd{a..z} first and not multipath devices.

Use pvscan -vvv to determine how lvm is viewing your volumes. You can also look in lvm.conf to help determine how LVM should be behaving.

On the "Urgent" matter. I was trying to point out that if a problem is that severe your best route would be to exercise your support contracts to get quicker support.