Long story short, I have a homebrew SAN publishing LUNs. I had that misconfigured a little bit and found that the ESXi host was connecting to these LUNs over a single, standard 1G/1500MTU connection instead of the 2 1G/9000MTU interfaces I had intended.
I have fixed that oversight on the SAN with target portal groups so now the LUNs only appear on the 2 IPs I want them to show on.
I would like to change the IPs of the iSCSI targets on the SAN so that the iSCSI traffic is segregated from the rest of my LAN traffic.
LAN IP=172.16.16.0/24
Proposed iSCSI net=10.0.0.0/24
Currently, I have the 2 interfaces on the SAN as 172.16.16.76 and .77 and the vmkernel IPs are 172.16.16.78 and .79 on a single vswitch. They are assigned to the iSCSI HBA using port groups and dynamic target IPs listed earlier.
This works fine.
When I change the IPs on the SAN and on the vmkernel ports, the host can ping the SAN and vice versa, the host can even see the presented LUNs, but performance is super slow. Also the LUNs don't show up as the datastores they were when they were on the other network. When I switched over to the other network (10.0.0.0), I removed the port group config from the iSCSI HBA and just did dynamic discovery. In fact I've tried it both with and w/o the port groups. Same results.
What am I overlooking here?
I was under the impression that a LUN is a LUN regardless of the IPs its bound to and presented from.
I have tried setting the 4 ports for iSCSI traffic on a different VLAN both with tagging and w/o, no help when using the new IPs.
I have even tried 2 separate networks on the SAN interfaces for iSCSI interfaces. Same result. Super slow performance.
I know that there is only 1 gateway defined because of the way ESXi console works, but even when I configured the other 2 vmkernel ports on a different network, that change reflected in the routing table and traffic would go out the correct vmkernel port as pings from the host to the SAN were successful.
Thoughts?
TIA
Sam