Automatically change vcpu's to allow failover to less powerful node

[ben90818532]

Hey folks,

I've got a small 3 node proxmox cluster, some nodes have different numbers of cpu cores.

This stops vm's assigned more vcpus than available on another node from failing over to the less powerful node.

Can we please have an option in the GUI to if necessary shutdown a vm that needs to failover and assign a compatible number of vcpu's before migrating?

It could be a percent of available cores on a node or simply a tick box "assign max cores".

Another option could be "emergency mode" in the HA configuration that automatically reduces the number of vcpu's to be compatible then raises them back to the previous setting when the main server is restored.

Two of the nodes in my cluster are significantly less powerful than the primary, only meant to keep things "moving but less quickly" if the main server goes down or fails.

Cheers

 

[n1nj4888]

This would be a good feature to add - The ability to configure a VM to use a Percentage of physical CPUs / vCPUs depending on the total available cores on host that is running the VM...

I also have a VM which is configured to use say 8 Cores but this VM cannot automatically start on my failover node which only has 4 Cores therefore I have to manually edit the proxmox configuration of the VM to reduce the configured cores...

Is there a way to formally request this feature from the Proxmox Dev Team?

 

[Stoiko Ivanov]

s there a way to formally request this feature from the Proxmox Dev Team?

The preferred way is to open an enhancement request at https://bugzilla.proxmox.com

hope this helps!

 

[n1nj4888]

The preferred way is to open an enhancement request at https://bugzilla.proxmox.com

hope this helps!

Thanks. I’ll raise the feature request and post the link back here!

Thanks!

 

[n1nj4888]

I’ve raised the Feature Request for this here...

https://bugzilla.proxmox.com/show_bug.cgi?id=2254

 

[RolandK]

why would ne need changing vcpu number? why can't we run a 8vcpu VM on a 4CPU node ? ok, it doesn't perform, but if it's an idle VM or if performance doesn't matter at night for example, what's the issue ?