HA, one nodte failure, vm reaction

[m4rek11]

Hello,

I have a Proxmox cluster built with 4 servers and HA configured. Now, a question that might sound silly: In the event of one server failing, all virtual machines are migrated to another server in line with the policy, which is obvious. However, will these virtual machines shouldbe restarted or retain the "running" status from before the migration? In my opinion, they will, but I’d rather double-check.

 

[Weehooey-HSS]

If I understand correctly, you are asking if VMs on the failed node will migrate to the other nodes while still running. If the node fails, the VMs do not get a chance to move the RAM over. They will effectively start fresh on the other nodes as if they were hard-powered off.

You can look at the shutdown policy here to change what happens with HA when a node gracefully shuts down.

You can also check out the HA simulator!

Side note! Add a Q device to avoid a split brain with four nodes!

 

[Red Squirrel]

The VMs are essentially shut down abruptly, like pulling the power cord, then after Proxmox realizes what just happened, they are restarted on another node. It's not as clean as I would have hoped but I guess to do it 100% live it would require that the ram state is somehow replicated in real time on another server and that would probably be really hard to do.