Today I had an interesting re-experience. What is a re-experience? A re-experience is problem/error/bug you had a while ago – but forgotten about. Then when it appears months or years later you scratch your head and wonder – hang-on what’s going on here.
For years I’ve warned people about the perils of renaming portgroups on Standard vSwitches. Firstly, their per-host. So any rename will have to be consistent across EVERY host that’s touched by that rename. Secondly, renames of portgroups can be dangerous because it can leave affected VMs orphaned from the switch. This shows itself in a couple of ways. On the properties of the affect VMs (and there could be more than one…) you would find the portgroup dimmed/italicized.
Here my F5 BIG-IP load-balancer has become disconnected from the portgroup. Here it happens to be an internal switch. I set up just to keep the management network of the BIG-IP private. It’s not a recommendation from F5 to do this incidentally. It was just a quick way of me isolating the appliance when I need to. If you edited the settings of VM like this – you would find that NIC interface lacking a portgroup reference (because of the rename) and also set not connect at power on. That’s pretty critical. Suppose an ESX host died and a HA event kicked in… Well, these orphaned VMs wouldn’t get powered on….
So where did my re-experience come in? Well, I was renaming portgroups (without engaging the brain, I should add). I couldn’t work out why the “Network View” in the vCenter – were showing portgroups that didn’t exist. I thought it was a “refresh” or “GUI” thing. Turns out the Network View kept hold of those old portgroup names until the VMs were mapped to the new name…
The portgroups highlighted in red – still had VMs “connected” them – despite the fact they didn’t exist on any host. Once the VMs had been mapped to the new networks – they disappeared.