topic Re: Power Path question in Operating System - Linux

Power Path question

Waqar Razi — Thu, 22 Jul 2010 15:27:38 GMT

We have RHEL 4.3 and power path installed on the server. I have just observed that in the VG which is on the SAN, they (implementation team) have used the pvlinks (dev/sdxx and only just one pvlink out of 3 and not all of them) instead of power path names (emcpowerx).

I have some questions here:

1- In this case, when we are just using the one pvlink and not all and not the power device names, are we still having the load balancing feature of power path?

2- Will it will failover to the alternate pvlink, incase if the one included in the VG fails?

3- How can we fix this issue? This is the production server and we cannot afford downtime or data loss. Please provide step by step method to resolve this issue.

Thanks for your help. Please find attached the output of vgdisplay -v and the powermt display dev=all commands output.

Re: Power Path question

Michal Kapalka (mikap) — Fri, 23 Jul 2010 06:32:51 GMT

hi,

i use the same config as you, i use physical device instead of pseudo device /dev/emcpowerX on Centos 5.x, without any problem, if i found some server with the FC connection, i will test the disk availability when the /dev/sdX used in LVM will fail simulateing the HBA failure.

In my personal opinion, it should work, but i never tested it, we are using PowerPath mostly on HP-UX and the is not some a pseudo device defined, you just see some LUN presented witn numbers of disks, and encapsulated together with powerpath.

mikap

Re: Power Path question

John Guster — Wed, 04 Aug 2010 17:02:57 GMT

it seems more likely the issue with the filter in lvm.conf file, can you post the output of grep -i filter /etc/lvm/lvm.conf?
Once the filter is set correctly, the vgdisplay will show the PP alias device name.