Evening all,
We've had a recent power outage that took out our Lab Manager infrastructure. Despite a full shutdown and staggered ordered reboot of the rig our user base has noted some odd behaviour on their deployments.
We're in the test and development game so use the fenced deployment options to allow us to roll out numerous matching instances of the same virtual subnet/domain design.
Admittedly the Lab Manager infrastructure crashed - a number of virtual port groups were left from the forced undeploy actions required on some of the configurations.
What appears to be happening now though is that the revert option is causing issues with networking and continues to do so even with new configurations.
Scenario.
1) Configuration deployed in fenced mode with a number of VMs in linked mode. VMs can ping each other on internal and external NAT address.
2) One of the VMs is powered off and snapshotted via the Lab Manager console and powered on again.
3) If the same VM is powered off again within the same deployment and reverted the VM continues to be able to ping other VMs in the deployment on internal and external addresses.
4) If the configuration is undeployed before step 3) takes place and redeployed with the same guest OS settings no changes at all and then is powered off and reverted the VM in question loses all connectivity with the external network. The VM guest OS sees a network connection, the NAT virtual router created with the deployment still pings the external network IP address it has assigned to the reverted VM as can the other VMs in the configuration. It's as if the connection for the VM has been patched into an invalid port on another switch somewhere.
The power failure did cause some VMs to report parentCID mismatches and fail to boot at all so reverting them to snapshots within a renewed deployment did work in those instances. My concern is that it also seems to be happening with new configurations I'm creating.
Can anyone help me work out if I should be looking at a virtual networking issue or this just a case of linked clone failure across the board please? I'm a little perplexed at the moment.
Thanks
EDIT - just to add to this - I deployed a test configuration in fenced mode - 2 linked clones and one full clone VMs from the same template. Within the deployment I snapped one of the linked and the full clone in powered off state. Undeployed and redeployed and all could still ping. Reverted the two snapped VMs. The two reversions can network with each other but on their internal addresses only and the unsnapped VM is is a network world of its own. All still appear correctly on the virtual port group on vCenter. Looking like the virtualrouter is an issue?