HA cluster....what`s the role

[Ivan Gersi]

I`m little confused/disappointed from HA manager in PVE. I tried to use DRBD (there was problem with IP via NFS if primary node failed), next iscsi(LVM) in double (active-active) NAS (this scenario is ok) but what`s the role of HA manager in PVE? Ok, it can be able to migrate machine from one node to another (changing priority e.g.) like a flash.

2019-11-19 19:16:38 migration speed: 910.22 MB/s - downtime 55 ms
2019-11-19 19:16:38 migration status: completed
2019-11-19 19:16:41 migration finished successfully (duration 00:00:15)
TASK OK

But the main role is failed...can I have HA machine if one node is failed? I don`t think so.
I tried to run VM in node4 a then restart this node (via shutdown -r now, not suddenly) and....node going down with my VM. When node was up my VM going up too. Why guys? What I didn`t understand? I hoped if node with highest priority failed, 2nd node take over VM online because its storage is available for every node (because iscsi storage).

 

[danielb]

When you reboot cleanly one node, the default behavior is to freeze it's resources (this can be configured, see man datacenter.cfg). HA in proxmox is made to recover from unexpected node issue. Try unplugging the network cable of a node, and you should see the node being self fenced, and it's HA resources beging recovered on other working nodes

 

[Seed]

Try unplugging the network cable of a node, and you should see the node being self fenced, and it's HA resources beging recovered on other working nodes

What if the node is unplugged from power or has a hard crash?

 

[tom]

What if the node is unplugged from power or has a hard crash?

HA will make sure that your VM/CT will start on the remaining nodes.

I recommend to play with our HA simulator, so you can learn and see how it works.

https://pve.proxmox.com/pve-docs/pve-admin-guide.html#_ha_simulator

 

[Seed]

HA will make sure that your VM/CT will start on the remaining nodes.

I recommend to play with our HA simulator, so you can learn and see how it works.

I can test this live. So if I shutdown one of my nodes with a VM on it that's set to HA for the whole group, I should expect it to appear automagically on another node after shutdown?

 

[tom]

I can test this live. So if I shutdown one of my nodes with a VM on it that's set to HA for the whole group, I should expect it to appear automagically on another node after shutdown?

Shutting down a node is not the same as a node crash. Start reading the docs for explanation.

 

[Seed]

Shutting down a node is not the same as a node crash. Start reading the docs for explanation.

ok "unplugged" then..... eesh.

 

[Ivan Gersi]

Yesterday at 20:56
Shutting down a node is not the same as a node crash. Start reading the docs for explanation.

I don`t understand why is this case different like power off or freeze, crash, etc because I`m lost 99.9999% HA in this case.
My logic is simple...the quorum decides and if node (rebooting, power off, lost network connecion...etc) lost connection from cluster/quorum it must stop the VM and rest of cluster must decides run VM from backuping node.

 

[danielb]

A "clean" reboot is not something unexpected. You can migrate resources before rebooting the node. So it won't affect HA at all. An unexpeced reboot, a crash, a non responding node is something completly different, and this is where the Proxmox HA stack will come to play, and recover resources. Please read https://pve.proxmox.com/wiki/Manual:_datacenter.cfg (especially the shutdown_policy section) to change the default behavior.

 

[Ivan Gersi]

Ok, now it more clear for me like before .