Storage Failover with LUN Promotion and RDM LUNs

[PhoenixHawk]

Hello.

I´m not really sure, if it´s the correct place for this discussion here, but I will give it a try.

Following environment:

Two mirrored EMC² Storage systems with Mirrorview. Four hosts, two per site. One Clariion is the primary, the other one is secondary.

Test scenario:

Disabled the primary storage, promoted the mirrors, checked the esxcfg-volume command output, rebooted the servers and all LUNs were successfully remounted and the VMs were available again.

NOW my actual question.

Up till now we only use VMFS volumes but it is planned to virtualize Oracle Servers with RDM mapped LUNs. What happens to these LUNs? Will these be remounted the same way and presented to the correct VMs or is there a manual procedure to do - e.g. reconfiguring the RDM mapping for each VM?

Would be glad to know, if there is anybody out here, who can help me. The trouth is in here, I know it.

[kjb007]

With VMFS, there is a signature on the LUN, that when re-presented, ESX can detect and allow you the option to force-mount a mirror LUN. This is not the case with RDM's. You could get lucky, and if your controller, target, device, lun, match exactly, the RDM may pick up correctly, but most likely, you will have to recreate the RDM.

-KjB

[PhoenixHawk]

That´s what I thought... We configured the Host LUN Ids exactly the same on both storages, so it might be possible, that the system recognizes the LUNs the exact same way but the storage underneath is changed, so most likely the RDMs have to be remapped. Thanx for you answer, I will check it out and post anything I find out in here.

[kjb007]

Remember the targets matter as well. It's difficult to make the targets show up as the same as the primary targets did, but again, you can get lucky.

-KjB

[binoche]

the storage underneath changed, especially its lun id naa.***, my guess you have to remap;

also I am not sure Oracle Servers can support re-signature or not, since lun id naa.*** changed, can Oracle Servers still recognize its data successfully?

binoche, VMware VCP, Cisco CCNA