I have a 1 TB virtual machine, which I cloned while shutdown and making the clone took 3.5 hours; which strikes me as a very long time. The source and destination are located on different datastores, but on the same EMC VNX5500 SAN.
I then converted the clone to a template, and have deployed 3 VMs from that template. Each of the deploys has taken just 1 hour; much more what I would expect. I've even made a cold clone of one of those VMs and it also only took 1 hour.
The original server is going to have its application upgraded, and then a fresh clone will be taken for a development system. The entire window for doing the upgrade and taking the clone is only 5 hours, and there isn't a way to fit that 3.5 hour clone in that.
So, I'm trying to figure out where the bottleneck might be when cloning the original VM. I am using Thick Eager Zero, but that is try throughout (i.e. the new VMs from the template, which clone in just an hour).
I've tried tracking down the actual cloning architecture to see if I'm hitting a situation where the data is passing through the vCenter server (possible network bottleneck), transfers between the 2 hosting ESXi hosts (potential network bottleneck again?), direct SAN copy (unlikely, but who knows), so something else I'm not even thinking about.
I'm also not sure where to even look for performance data while the cloning is occurring.
Unfortunately, taking a fresh clone of the source and doing the upgrade on that, or using one of the new new VMs as a replacement isn't viable. Data on the server will be changing up to the very beginning of the 5 hour upgrade window.