VM falls back to node1

[tjarcoboerkoel]

Dear readers,

I have a 3 node cluster on iSCSI storage
- node1
- node2
- node3

My VM (ubuntu) was running on node1 and the network link failed so proxmox tried to migrate it to node2. node2 has to few cores so there failed the VM to start.
Node1 is back online and I performed the ha-manager set vm:102 --state disabled, check everything to be oke and then ha-manager set vm:102 --state enabled.

VM error state is resolved and I try to move the VM to node1. Proxmox acked with migration task... OK.

Weird is that Proxmox starts a new migration back to node2!? And again it fails on node2 because of to few cores. I don't understand why, node1 is running without issues.

If I change the CPU config for the VM to just 2 cores the VM does start without problems.
I then can perform a migrate to node1 (live), and after some minutes it start migrating again back to node2!?

Best regards.

 

[Chris]

Dear readers,

I have a 3 node cluster on iSCSI storage
- node1
- node2
- node3

My VM (ubuntu) was running on node1 and the network link failed so proxmox tried to migrate it to node2. node2 has to few cores so there failed the VM to start.
Node1 is back online and I performed the ha-manager set vm:102 --state disabled, check everything to be oke and then ha-manager set vm:102 --state enabled.

VM error state is resolved and I try to move the VM to node1. Proxmox acked with migration task... OK.

Weird is that Proxmox starts a new migration back to node2!? And again it fails on node2 because of to few cores. I don't understand why, node1 is running without issues.

If I change the CPU config for the VM to just 2 cores the VM does start without problems.
I then can perform a migrate to node1 (live), and after some minutes it start migrating again back to node2!?

Best regards.

Hi,
can you verify that the cluster is healthy and you did not loose quorum again? What is the status of pvecm status?

Check the systemd journal on both nodes for errors and further hints on why the VM migration is initiated. You can get a paginated view of the journal in reverse by running journalctl -r and if you would like to share it here you can get the journal since boot by journalctl -b > journal.txt.

 

[tjarcoboerkoel]

Hi Chris,

Thank you for your reply.
I did something with the group, allowed on all nodes, and now it does run on node1.
I'm not sure what happened here, the group was allowed to run on node1 and node2. Something seems to be reset.

Best regards,

 

[Chris]

Maybe no quorum while editing the configs? You will have to provide the requested outputs for us to help.

I did something with the group, allowed on all nodes, and now it does run on node1.

Please share these as well by posting the output of cat /etc/pve/ha/groups.cfg.