I have two IBM DS4700 storages with ERM (enhanced remote mirroring) enabled replicating LUNs synchronously. There are 85 VMs on ESX4/vCenter4 environment. On VI3, once one storage fails (the one that is read/write) all VMs went down and I had to change the second storage to primary (i.e.: from read only to read/write) and then remove all VMs from the vCenter inventory and import it again (from the “new” LUN discovered after a rescan on FC HBAs) using datastore browser tool.
Since VMware redesigned storage Access layer on vSphere4, does anyone have an idea if now there is no need to rescan LUNs and reimport VMs after the secondary storage takeover??
Does anyone have the same environment and wants to share some knowledge??
I have done some experiments months ago in a lab that is probably similar to your production setup:
http://it20.info/blogs/main/archive/2009/07/04/243.aspx
my sw stack was 3.5 because that is what SRM supports as of today so I can't say whether this will change or not with vSphere.
My take is that all the processes described in the article should remain pretty much the same with vSphere (i.e. I don't think you will be able to skip the "rescan" operation as that is a fundamental step to make aware the hypervisor that "new" storage is available).
This is pure speculation off the top of my head.
Massimo.
Not sure but would Vmware Site Recovery help at all with the rescan problem?
Ohps I apologize.
I went too quickly through the first post and I thought it was a typical DR situation. Apparently it seems to be a single site config (campus-like) with two mirrored storages (for Storage HA).
While one of the many key things that SRM does is the rescan of the surviving SAN... in such a context SRM would not even work (as it requires two different datacenters with two distinct vCenter servers etc etc).
Forget SRM then.
Having this said I doubt vSphere has changed anything with regards to SAN failover automation.
Massimo.