HA reboot node1 (Master)

[sand-max]

Hi guys!

I have 4 node in cluster. Now 2nd node is down. I created HA group to node3, node4 for test.
Added VM(node4) to group , and now when I reboot node 4 VM does not migrate, because Qourum is not OK.
But when I reboot node3 or node4 , node1 also reboot itself.

in log node1:

Jul 10 13:59:34 pvenode1 cron[4750]: (CRON) INFO (pidfile fd = 3)
Jul 10 13:59:34 pvenode1 cron[4750]: (CRON) INFO (Running @reboot jobs)

 

[mira]

See https://pve.proxmox.com/pve-docs/pve-admin-guide.html#chapter_ha_manager
If there are resources defined for a node and it does not have quorum, the watchdog won't be reset and the node gets fenced. This is by design.
So make sure you have quorum(in your case at least 3 nodes) at all times on nodes with HA resources defined.

 

[sand-max]

See https://pve.proxmox.com/pve-docs/pve-admin-guide.html#chapter_ha_manager
If there are resources defined for a node and it does not have quorum, the watchdog won't be reset and the node gets fenced. This is by design.
So make sure you have quorum(in your case at least 3 nodes) at all times on nodes with HA resources defined.

Yes, I've got it. I suppose there are 4nodes and if 2 nodes are down. Quorum is not OK so VM will not migrate to other nodes of cluster
But why if I reboot node4 or node3, node1 reboots also?
I found similar issue https://forum.proxmox.com/threads/ha-cause-the-reboot-of-node-if-you-reboot-a-master.26212/

 

[mira]

This is explained in the docs. If there are HA resources defined on the node it will reboot if quorum is lost for more than ~60 seconds. In your case you have a HA resource defined on node1 and once node2 and one of node3 or node4 goes down, you lose quorum. When this happens node1 can't reset the watchdog and it will get fenced.

 

[sand-max]

This is explained in the docs. If there are HA resources defined on the node it will reboot if quorum is lost for more than ~60 seconds. In your case you have a HA resource defined on node1 and once node2 and one of node3 or node4 goes down, you lose quorum. When this happens node1 can't reset the watchdog and it will get fenced.

Thanks!
Also I found one more post https://forum.proxmox.com/threads/proxmox-5-3-6-ha-setup-node-reboot.50184/

I've tested cluster of 4 nodes on Virtualbox. And noticed:
1. Really in that case nodes is reboot which have status in lrm "Active". If node is "Idle" it will not be reboot.
2. For example. If you have VMs in HA on node-4 and migrate it to node-3(node will change status to "Active") then migrate back to node-4. Node-3 will not change status "Active" to "Idle" and will be reboot in case lose Quorum.