vMotion issue since applying ESXi 5.0 Update1

[preecep]

We have a simple ESXi 5.0 two node cluster.

- Both hosts are HP DL360 G7 servers with 6 nics. We have 3 vswitches in place, each has two physical nics assigned/bound:

.- - - vSwitch0 (vmnic0 & vmnic1) = VM Network and Management Network on 100.200.1.x/24

- - - vSwitch1 (vmnic2 & vmnic3) = iSCSI traffic on 192.168.100.x/24
- - - vSwitch2 (vmnic4 & vmnic5) = vMotion on 192.168.100.x/24

Our production network resides on VLAN1 using the 100.200.1.x subnet. VLAN2 (ie: the 192.168.100.x network) is used for iscsi/vmotion traffic only. All nics are connected to a stack of HP ProCurve gigabit switches. Naturally both VLANs are configured on the stack.

Since updated both hosts with ESXi 5.0 Update 1 we are no longer able to live migrate vm's between hosts. We can however, power off the vm's and cold migrate vm's between hosts. If we disable vmotion on vSwitch2 and enable vmotion on vswitch0 it works fine. All physical nics are the same make/model hardware and all using the same bnx2 driver. Other than applying U1 nothing has changed on the storage or network infrastructure. All nics connect to their relevant VLAN'd ports on the same HP ProCurves. We logged a call with VMWare but they've deemed this as a network issue and have passed the buck back to us. Anyone seen anything like this before?

Any advice or suggestions would be welcome.

[a_p_ · Accepted solution]

I can't tell you for sure whether this is related to the network configuration, but I would definitely not run iSCSI and vMotion in the same subnet. Why don't you change the vMotion subnet to 192.168.101.x (unless this is already in use for something else) and then try to vmkping the vMotion IP addresses between the two hosts?

André

[preecep]

Wow... I have to say when I read your reply this morning I was a little sceptical that would resolve the issue. But hat’s off to you my friend, it worked. I deleted and recreated vSwitch2 (using a different subnet) on both hosts and surely enough we could live migrate once again.  I had assumed the issue was with the physical switches and/or a VMWare nic driver issue. André, many thanks you have just saved me a lot of time and pain, thank you.