[Help] Noob question how I can put my server in S3 suspend?

[Gsmalla Idris]

I have supermicro-x10dai that support s3 sleep mode when I run

systemctl suspend

what's happing is my system goes to deep sleep but it restarts itself instead of resuming from the last state.

AMD cards have this rest bug and putting the system in sleep mode and resume it will make the card available for the guest again without restarting the whole machine.
Waiting patiently for your suggestions.

 

[LnxBil]

Waiting patiently for your suggestions.

I doubt there will be many. I've never heard or seen someone that puts a server into sleep mode. What should be the point of that in a server environment?

 

[Gsmalla Idris]

I doubt there will be many. I've never heard or seen someone that puts a server into sleep mode. What should be the point of that in a server environment?

I use the server as a workstation with multiple os running as VMS and some times I wanna just put it in low power mode and the second reason because the AMD rest bug.

Here in this video a better description of the problem I have with AMD and how sleep will help to fix it.


Is there any simple way to achieve that by some scripts or by editing sleep config in /sys/power/state?

 

[masgo]

I doubt there will be many. I've never heard or seen someone that puts a server into sleep mode. What should be the point of that in a server environment?

It would make sense in a home-server environment. There, the server does not need to run 24/7. I do the same with my backup NAS. It just WoL it, do Backups and it goes to sleep by itself 1 hour after the last connection was terminated.

 

[fiveangle]

Came here looking for the same. We have a production cluster of new efficienct lean machines, but our Proxmox Backup Server is running on quite old/inefficient repurposed hardware with 4x fully populated 16-bay JBOD SAS trays added running ZFS on PBS (these are also repurposed relatively small 2TB WD SAS drives which we had hundreds of from a previous decommission). We require only nightly backups and it just doesn't make sense to have nearly 2 killowatts of systems running all day for no reason. We also like the benefit of fast rebuild times with the smaller RAIDZ2 member drives. We don't have the heat load capacity in the current micro server room to handle it so we are currently doing a nutty combination of scripting and network-controlled power switches to get around the problem, but it's been problematic at best.

It would be so much more ideal to be able to just sleep the proxmox server (we have PBS virtualized on it's own PM cluster) then hit the nodes with a magic packet as part of the backup process.

I'll keep looking, but anyone else doing the same, please post back here if you ever find a reliable solution.

Thanks !