I've been testing failure modes on an EQL PS4000XV. At some point paths to storage in ESX have got confused and show either one or two connections to the same EQL IP address for a LUN, rather than one connection to each of it's two addresses.
I've tested controller failover and pulling interconnecting cables and a switch, so am unsure what has caused it.
Question is how to get them back to what they should be. Re-scan doesn't do it and sequential performance and EQL monitor confirms only a single interface is being used for affected LUNs.
General info on setup:
- PS4000XV array (two GigE NICs)
- Two Force10 switches with 4x GigE LAG between them
- Three ESX hosts, build 208167
- Two NICs for iSCSI on one vSwitch
- Two vmkernals for iSCSI on same subnet
- LUN policy set to round-robin
- nmp roundrobin IOPS set to 3
- Servers and EQL have one connection to each switch