LM ver. 3.0.1.378
VC ver. 2.5.0 Bld 119598
ESX ver. 3.5.0 Bld 163429
We just recently remediated a good amount of updates against our LM cluster's ESX hosts. After the maintenance, we're noticing semi-sporadic issues with LM deploying fenced configurations. During a deployment, I can see that it attempts to create a virtual switch that already exists, which fails, but then doesn't seem to put the VM's into the port group of the already existing vSwitch.However the VirtualRouters are put into the correct Physical Network's vSwitch. The solution simply has been to undeploy/redeploy the config, works every time.
We've been working like this for a few days now and it's becoming more of an impact than originally estimated. Now when I go into vCenter and examine the ESX host's Networking Configuration tab, I am now seeing roughly 25 empty vSwitches across 3 of the 5 hosts that are currently in the cluster. Manually removing them doesn't seem to help. The leftover vSwitches are a mixture of both Physical Networks and Virtual Network Templates that we've built up.
Has anyone ever seen this? Is this something resolved in LM4.0?
Thank you in advance.
Patrick
Tags:
leftover_virtual_switch,
fencing,
network