So previously, running vSphere 6, configured with MPIO (each host has 2 controllers to SAN, 1 to controller a, and 1 to controller b), using round robin (of course both links are active, but only 1 serves I/O because of disk group ownership).
Upgraded to vSphere 6.5, and more recently, I migrated from linear to virtual disk groups (deleted linear, created virtual, migrated data back).
Anyways, all is good. However I'm noticing one thing while doing failover tests, not sure if it's the vSphere upgrade, or the change from linar to virtual. I'm thinking it's the switch to virtual, but would like confirmation.
Previously with MPIO, disk group ownership, there is a prefered path, however if the iSCSI network connection went down, the other controller would service I/O, and the datastore would remain functioning (it would fall over to the non-owned controller).
However now, ESXi still reports both paths as active and valid, however if I pull the network cable (or disable a path), the I/O doesn't switch to the other controller.
I'm wondering if by switching to virtual, the entire controller needs to go offline for the other controller to service I/O to the disk group?
Cheers