Below is a fairly detailed description of the issue. Any help/feed back would be great.
Basically the situation is that even after we do a complete failover vCenter is not allowing us to mount the VMFS datastores on all hosts because we are keeping the existing VMFS signature. We have to do this otherwise we will break the paths for all secondary vmdk's. The issue we are seeing is documented in Vmware KB 1015986: http://kb.vmware.com/kb/1015986. vCenter sees this as a snapshotted "replica" LUN even though it is not. We can force mount the VMFS datastore on the other ESX hosts and then we can Vmotion and everything works fine. Eric even wrote a nifty script to handle this.
Again, the EMC SR# is 45306736. Here are my notes I put in on the case answering the engineer's questions:
1) Are you performing the RP replica access test on single LUN ( one datastore) ?
We are trying it on two LUNs/datastores that are being used for testing purposes only.
2) When you referring to only one ESX host successful mounted the RP replica, do you means the always the same ESX host manage to mount the RP replica ? or do you means the RP replica can be mounted to any ONE ESX hosts at any given time ?
The LUN at the secondary site, when image access is enabled, can only be mounted on one ESX host at a time through vCenter. We are keeping the existing VMFS signature to preserve paths to vmdks. If we use a script we can force mount it on other ESX hosts but the customer would prefer not to have to do this.
3) Can you detect the replica LUN on ESX hosts (not mounting the datastore) ?
Yes, the LUN is seen by the FC HBA's on all ESX hosts.
4) Are ESX hosts at DR site are part of same VMware cluster in Production (means stretched VMware cluster) ? or ESX hosts at DR site form a separate VMware cluster from production cluster ?
They are in a different VMware ESX cluster at the DR site and are using a separate vCenter server as well.