So uh, I'm sure I know why this happened, but here goes.
I was having some communication issues on one node after a disaster recovery and reinstall. I checked all the config, but VLANs on this VM just wouldn't communicate outside the host. So I decided to attempt to reboot the switch. (Note, this is a home office dev & prod environment, so no redundant LAN yet). When I did, I noticed that one of my pve hosts was also rebooting, so I plugged the monitor into the others, and they were *all* power cycling.
So here's my hypothesis. I think each node thought it lost contact with the other PVE nodes in the cluster, thought it was in a fault state, and attempted to reboot itself to resolve the issue. A watchdog of some sort. A sane default most of the time. The thing is, I didn't set that behaviour up, and I wouldn't want it to happen in the future if I reset the switch for any other reasons in the future.
The questions: Is my assumption correct? Is there any way I can permanently, or preferably, temporarily, disable this feature, short of providing a secondary redundant cluster link?
I was having some communication issues on one node after a disaster recovery and reinstall. I checked all the config, but VLANs on this VM just wouldn't communicate outside the host. So I decided to attempt to reboot the switch. (Note, this is a home office dev & prod environment, so no redundant LAN yet). When I did, I noticed that one of my pve hosts was also rebooting, so I plugged the monitor into the others, and they were *all* power cycling.
So here's my hypothesis. I think each node thought it lost contact with the other PVE nodes in the cluster, thought it was in a fault state, and attempted to reboot itself to resolve the issue. A watchdog of some sort. A sane default most of the time. The thing is, I didn't set that behaviour up, and I wouldn't want it to happen in the future if I reset the switch for any other reasons in the future.
The questions: Is my assumption correct? Is there any way I can permanently, or preferably, temporarily, disable this feature, short of providing a secondary redundant cluster link?
