[SOLVED] Correct way to restart a node without stopping containers or vm's

[lifeboy]

Is there a correct way to restart a node (mostly for applying kernel updates)?

I have about 20 virtual machines and containers that should not go down ever unless planned. So I have set the all as HA. When I shutdown a node, the VM's / LXC's shut down instead of migrating to the node next in the HA config. I can of course reset the node, but then it's like crashing the node, which is not what I want to do.

How does one ensure the HA config is honoured when a node is shut down and restarted?

 

[t.lamprecht]

So I have set the all as HA. When I shutdown a node, the VM's / LXC's shut down instead of migrating to the node next in the HA config. I can of course reset the node, but then it's like crashing the node, which is not what I want to do.

Change the "Shutdown Policy" to "migrate". This can be done in the webinterface under Datacenter → Options → HA Settings.

 

[lifeboy]

That does migrate those LXC's or VM's set up with HA. However, is there a way to get other running VM's or LXC's to hibernate and be restored to their running state when a node is restarted, rather than to have them shutdown? That would be really useful.

 

[t.lamprecht]

For LXC CT, no, that's not possible and won't be anytime soon.
For VMs it can be done manually, but not yet automatically - this is planed though.

 

[lifeboy]

For VMs, yes, I haven't noticed that I can hibernate the machine manually before restarting a node.

Most LXC's are services, so it's not that relevant. I just realised that it's mainly machines where users log in remotely to where this would be useful so that users don't lose their files and session contexts.