We had a failing iSCSI array (MD3220i) so I began the process of configuring a replacement array (also MD3220i) with identical disk pools, etc.
As we migrated VM's from the old datastore to the new, we noticed poor performance and I began to look through my configuration. My old path settings were set to Round Robin (all four paths state "Active (I/O)", while my new array appears to have been configured as MRU (all four paths show "Active", only one shows "Active (I/O)".
Question: Should the VM's be off and host in maintenance mode before choosing Round Robin as the path selection?
Thanks in advance,
There should be no issue with switching the path policy to Round Robin. I've done this many times myself.
In addition to switching to Round Robin, I'd suggest that you also check other settings like "DelayedAck" in the iSCSI settings, which can cause high latency for arrays, which do not support this.