MD3600i - dual controller, 2 10Gbps iSCSI ports per controller, jumbo frames enabled on switch and array
2 MD1200's and 1 MD1220 both with dual EMM
3 RAID 10 virtual disks
2 RAID 6 virtual disks
out of warranty
VD's are mapped to hyper-v 6 node cluster and nothing has changed since this storage was implemented five years ago.
Ownership/preferred path has 3 VD's on controller 0 and 2 VD's on controller 1. Five months ago, we got the warning "virtual disk not on preferred path due to failover." The hyperv owner node was rebooted and the problem went away. The problem returned and rebooting the nodes didn't help so the recovery guru steps were followed. It appears that iscsi connections between the nodes (2 each), the two stacked switches, and the two MD3600i controllers are all linked solid with jumbo frames enabled throughout. There are two VLANS for the iSCSI networks. Cabling and switch ports have been changed.
Redistribute virtual disks and they go back to their preferred path on controller 1 but failover to controller 0 after exactly 10 minutes. At 5 minutes, there is an informational event "Virtual disk I/O shipping implicit transfer" then the failover at 10 minutes. If there is no activity on the VD's, they maintain their path on controller 1 indefinitely. Does this sound like the controller card needs to be replaced?