EVA8100, CA, VMWare 3.5 and RDM disks

[Adam Garsha]

In this link:

http://h71028.www7.hp.com/ERC/downloads/4AA1-0820ENW.pdf

I see for VMFS between same array controller type, CA failover is a breeze since VMWare uses WWID/Controller-type/LUN-ID meta data only to determine uniqueness of a SAN disk. Excellent!

So I am confident how to go about migrating a cluster from a VMFS standpoint (and will test it). My problem is with RDM disks.

In the link it says:

"RDMs are referenced differently than VMDKs. In the event of a failover to the storage system in the other data center, the path to all RDM LUNs will have to be manually re-configured into each VM configuration."

If we are talking about strictly windows server VM's does this hold true? Can anyone provide details or insight into the process when RDMs are involved.

The model I will be following is:
Replicate all VMFS and RDM LUNs using CA to another like-type Array and set LUN IDs to be exactly the same. Failover. Depresent original source luns. Break dr_group relationship and delete original source luns when I am sure all-is-cool.

The thing I don't know is what is impact on RDM LUNs (we do physical and we keep the mapping file with the VM).

Thanks!

[Adam Garsha]

The special notes section on RDM gets me close:

http://www.virtualization-symposia2008.lv/files/File/3_vmware_virtualization_symposia_latvia_thuber.ppt

But not sure what effects "LUN_ID" (vs. LUN#)... fear that it is based on order of presentation which would force remove/re-add of RDM on all VM's that own an RDM.

Any insight, good insight. I hope to test in coming days a whole-sale controlled data move of esx cluster to a different but like typed eva.

[Adam Garsha]

... but if by lun_id in the slideshow they mean "wwid" with EVA's I am golden/good.

[Uwe Zessin · Accepted solution]

I _love_ the term "LUN ID" :-(

Looked at some slides and "LUN #" means the "LUN address" assigned during the virtual disk presentation (1..255). But on slide 21, they get confused again, because in the VMware world, the term "LUN ID" __usually__ means "LUN address", too.

"WW LUN ID" is the "LUN WWN" which by default is taken from the EVA's internal UUID - a 128-bit value which starts with "6". On slide 11 this is also called the "NAA ID" - see, even one presentation is not consistent!


If your tests with the default parameters are not successful, you can try a workaround AS LONG AS YOU DO NOT MAP ANY SNAPSHOTS to your ESX servers:

[root@esxc root]# esxcfg-advcfg -g /LVM/DisallowSnapshotLun
Value of DisallowSnapshotLun is 1
[root@esxc root]#

Set the value to 0, recan and see if you now get at the data. Before you do that, see the "SAN System Design and Deployment Guide":
http://www.vmware.com/pdf/vi3_san_design_deploy.pdf


I suggest that you save the "LUN WWN" from the source array. If the value somehow gets lost on the destination array, you can alter its vdisk's "LUN WWN"s as long as they are unpresented.