We're in the process of setting up a 2 node ESXi 4.1 cluster. For redundancy we purchased two iSCSI SANs (left-hand HP P4300) and two HP Procurve switches. This idea was we could have one switch, one SAN or one ESXi host fail without causing any downtime.
We installed a HP vApp that participates in the HP SAN management group and controls the quorum. This appears to be working well as we can power off one of the SANs without this effecting the ESX hosts or vm's.
However, when we power off one of the HP Procurve switches we lost connectivity to our VMs until both switches are brought back online.
The ESXi hosts have 6 physical nics which we've paired into 3 vSwitches, see screen shot from one of the hosts below (both are setup identically):
- VM Network
- Management Network
- iSCSI traffic
- iSCSI VM Network (only used to allow a vApp and vCentre vm operate on the iSCSI lan)
Note: Here we created two vmkernal ports and linked each to a separate physical nic, as per vmwares best practice.
- vMotion traffic
This is the HP white paper we've followed. Does anyone have an ideas where we went wrong?