We performed an SRM failover/failback between two sites (successfully). On the recovery site we notice that the replicated luns that were mounted during failover, now show as unmounted during failback. Should SRM have followed that unmount with a detach, so no devices would be showing on the recovery side hosts?
I ask because we recently patched the hosts at the recovery site, which meant they got rebooted. After reboot, the previously unmounted luns are now showing as mounted. To be more specific, they are simply mounted (connected) to the ESXi hosts at the recovery side. The actual datastores associated with those luns have not been mounted.
Is this the behaviour of SRM after a failback? Should we have expected to see the recovery hosts with no evidence of those replicated luns (i.e. SRM should have unmounted and then detached the luns)? I'm not even certain that would suffice as the lun mapping at the storage side still exists. I suppose a full unmount would be unmount - detach - unmap. I don't believe SRM goes that far when unpresenting the replicated datastores from the recovery hosts during a failback. Any clarification would be appreciated.
I believe this may vary depending on which storage replication technology is in use. As I recall, certain solutions (Recoverpoint maybe?) do not fully detach the LUNs from the hosts, but with other vendors the storage is completely un-presented as part of the failover process.
Hi p_hall. Thanks for the feedback. You would think that since VMware vets each storage vendors SRA, they would demand a common framework. It's just odd that they wouldn't demand each storage vendor completely unpresent the storage after failover/failback. The storage vendor we're using is Hitachi VSP. I can't find anywhere that confirms what they do.