Hello!
Recently I've been playing with various vCenter and PSC failure & recovery scenarios in one of VMware recommended topologies (https://kb.vmware.com/kb/2108548).
More specifically, the one on the picture:
Turns out, failure of any node makes management of the system quite hard due to Web Client behaviour.
Issue is:
- All vCenters are in Enhanced Linked Mode
- In case of vCenter's failure (PSC's failure counts as failure of all vCenters connected to it), attempt to login to Web Client on any surving vCenter spams /var/log/vmware/vsphere-client/logs/vsphere_client_virgo.log with messages like "Unable to connect to VMware Inventory Service - https://vcvdia1.corp.vmw:443/invsvc"
- You could think Web Client will show inventory of suriving vCenters, but actually it is mostly hangs forever doing endless queries to dead comrades and is mostly useless
Sometimes you could work around it by rebooting one of the surviving vCenters (but i guess restarting just inventory service would do) and reconnecting to it's Web Client. But it worked for me in, like, 30% of cases, and reboot takes so much time!
Unregistering failed vCenter from SSO (https://kb.vmware.com/kb/2106736) is not a solution in my opinion, as it is permanent and you cannot reconnect failed vCenter to SSO domain after it's recovery.
While it is fun to play with all of this in lab, I wouldn't like to face this issue in production.
So, I'd be grateful if someone would share solution or workaround for this issue.