Hi,
We are in the process of upgrading our farm from vSphere 4.1 to 5.1. The first part of this process was to upgrade the vCenter Server to 5.1 and then assuming all was ok we were then planning on upgrading the hosts to ESXi 5.1 - currently they are ESX 4.1 Update 2
Since the upgrade of the vCenter Server however, we've noticed that the snapshot creation process is now taking a huge amount of time to complete. The process of removing the snapshot is still very quick.
From a testing point of view, this is what we've done.
All snapshot jobs are including the virtual machine's memory. Removing the option 'Snapshot the virtual machine's memory' results in the process taking a couple of seconds.
We've also upgraded one of the hosts to ESXi 5.1, moved some VM's to that host and tried the same snapshot process but the fault remains. For a VM that is on the new host, we've also upgraded VMware Tools which produced the same result and then upgrade the hardware version of the VM to 9 - but again produced the same results.
Connecting to an individual ESX 4.1 host and then trying to create a snapshot from there also produces the same issue. Same with connecting to the individual ESXi 5.1 host.
If the VM is currently on an ESX 4.1 host, the status jumps almost instantly to 95% and then waits there until it's complete. On an ESXi 5.1 host, the status slowly makes it way up to 100% consuming the same sort of time. The amount of time it takes to create a snapshot varies per VM which I'm assuming is down the amount of RAM it has allocated to it.
An example of VM which has 16GB RAM
Prior to the vCenter upgrade, creating a snapshot took about 20 seconds - some were a little quicker depending on the time of day. That particular VM uses quiescing as part of the options as well so it's expected to take a little longer. Since the upgrade, creating a snapshot now takes almost 20 minutes. We have one VM with 64GB RAM and I can only assume this would take more than an hour to complete.
The above is just an example. The issue appears to be for all of our VM's regardless of OS, location on datastore, network profile etc. The process also causes network interfaces to drop momentarily which causes outages to the various services the VM is providing which is a major issue for us. We also use snapshoting for replication so all in all quite a significant problem.
Anyone else experiencing this problem?
Thanks.