HA setup and reboot due watchdog

[beanyzz]

Hello,

We are running a proxmox VE cluster and we where planning to use HA for some vm's, but not for all.
Due to the design of HA I understand a node would reboot after the watchdog timer expires. And it looks like this cause vm's not participating in HA to be unreachable on this node.

If I for example have a cluster of 10, and assign 2 nodes in HA group.
- I assume I need to use only HA vm's on those 2 nodes?
- Will the watchdog on the other 8 nodes not in HA reboots servers when the watchdog has issues?
- How long will it take before the HA vm will start on the other node, because I observed a case where I have 3 nodes in a group, but the vm was not started on another node. This can be miss config, but I would like to know what to expect.

Thanks in advance.

 

[brucexx]

That has not been my experience but I am on 6.4-1 version. I have mixed VMs on a cluster participating. Majority of VMs participate in HA but small number of them do not. At least on 6.4-1 it is not a problem and you can create and run VMs not participating in HA on that node.

Thx

 

[brucexx]

To answer your other questions at least as 6.4.-1 is concerned, I think that the HA is a cluster and not a node setting. You setup it up under Datacenter. Then you tell the datacenter which VMs are participating and which are not. Perhaps you setup the state of the VM to something other then started ?

I had a node reboot and stuck on boot due to an excessive ECC unrecoverable errors. The VMs that participated in HA simply started right away on the other node but the VMs that did not participate stayed on the failed node.

 

[beanyzz]

That has not been my experience but I am on 6.4-1 version. I have mixed VMs on a cluster participating. Majority of VMs participate in HA but small number of them do not. At least on 6.4-1 it is not a problem and you can create and run VMs not participating in HA on that node.

Thx

Yes it is possible, but the fencing thing reboots the node when it beleeves there is something wrong with it.
https://forum.proxmox.com/threads/proxmox-cluster-disable-watchdog-restart.121065/post-526481

But my question was that if hosts are not added to a HA group if the fencing is still active on the other nodes in the cluster not in a HA group? Or is fencing / the watchdog reboot only active when the node belongs to a group? But maybe @sterzy can answer this question?

I'm running 7.2-x since i'm migrating from 6.4

After I checked all my hardware I will do some testing in a maintenance window to see the failover in the HA group.
Not sure why the vm in the HA group did not failover when it rebooted the node due fencing.