EnableResignature and DisallowSnapshotLun values after failover

[galday]

Hi there, one of our customers had a problem with some DataStores suddenly recognized as Snapshots.

The customer has an active-active setup of SRM between 2 sites, with recovery plans both ways (tested OK).

We filled an SR and after too many questions and answers, we (and not support) discovered the values of /LVM/EnableResignature and /LVM/DisallowSnapshotLun were different from defaults (EnableResignature 0, DisallowSnapshotLun 1) in all of the customer ESX servers.

Support has answered this after asking them about this:

SRM will change those values on the recovery site and they will need to be set back manually but they should not be getting changed on your protected site. This would occur on a test recovery as well as a real recovery.

We have searched through the docs and we can't find anywhere such advice. Strange as it seems quite significant if true as every SRM environment gets messed values and can get into trouble with LUNs easily.

Did anyone of you know about that?? What do you think?

Thanks in advance.

[galday]

Message was edited by: galday

[scasg]

I have a site configured with SRM and can confirm that this very nasty little bug is there. It caused me untold pain after I powered a server off (it wasn't an issue with a simple reboot) and brought it up again. Luns got mounted then resignatured and caused all my VMs to fail. Fortunately it was the DR site. We managed to recover the situation, but only after hours of work. I am about to raise this with VMware as I have just performed a test to confirm the problem after being told it wasn't possible. Note this is with an HP EVA SAN.

[TimOudin]

Everything you describe in your first post makes perfect sense. The EnableResignature and DisallowSnapshotLun values are indeed changed during the installation processes SRM. This ability to resignature LUN is necessary for SRM. Unfortunately, in VC3 enabling these options is a host global thing but has been changed in vSphere. These are not optional parameters, this is not a bug and don't try to recover to production ESX hosts for the ultimate in safety. Your production LUNs should not be resignatured in this process, should only be performed on LUN recognized as a snapshot...IIRC.

Again, this gets much better in vSphere. Wish I had a better response for your current situation.

Tim Oudin, not a VMware engineer and speaking with no authority