The preference is that they be the same, improves compatibility with some hosts.
Pain though it is to do in practice, especially if you already used LUNs from both arrays.
I consider it a tidier method for diagnostics and less confusing for host admins.
I believe VMware can cope with them being different but it might need an option setting in the ESX consoles to enable.
So if you can, match them, if a lot of pain involved try without but test each OS to make sure they survive a switchover (on a test LUN
). You can look at the paths on the host to see if it lists both the active and standby ones, can also get into digging deeper into the TDID numbers on each path.