hibernate a VM via qm command

[tincboy]

I didn't find proper qm command options to hibernate or resume a VM via terminal,
Would you please let me know what are the options if any?

 

[sb-jw]

qm suspend <vmid> [OPTIONS]

Suspend virtual machine.

qm resume <vmid> [OPTIONS]

Resume virtual machine.

https://pve.proxmox.com/pve-docs/qm.1.html

 

[tincboy]

qm suspend doesn't hibernate the VM, it just halt it.

 

[Stoiko Ivanov]

check the manpage of `qm` - for hibernation you use `qm suspend --todisk 1`

hope this helps!

 

[masgo]

Does qm suspend --todisk 1 also work if the pve node is rebooted afterwards? So: suspend VM, reboot node, resume VM. I found this bug report https://bugzilla.proxmox.com/show_bug.cgi?id=426 where it looks like this would work, but I am not completley sure if it really applies to this case.

 

[davidand]

@masgo Yes, I think it does. I did:

1) qm suspend 100 --todisk 1
2) rebooted PVE
3) VM started automatically
- lastlog from the VM showed last reboot about a month ago

Therefore I think yes, it works like you described.

 

[Jannoke]

For me the confusing part is that where is the hibernation data kept? We got some guest machines with 250GB+ ram, so where will this all pushed?

 

[davidand]

@Jannoke Although I don't have any exact knowledge of this, this might be useful from the documentation:

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

State storage selection
If no target storage for the memory is given, it will be automatically chosen, the first of:

1. The storage vmstatestorage from the VM config.
2. The first shared storage from any VM disk.
3. The first non-shared storage from any VM disk.
4. The storage local as a fallback.

 

[davidand]

@Jannoke I did a small test and for me it did create a new volume:

root@proxmox:~# qm suspend 103 --todisk 1
Logical volume "vm-103-state-suspend-2020-05-10" created.



root@proxmox:~# qm resume 103
Resuming suspended VM
activating and using 'local-lvm:vm-103-state-suspend-2020-05-10' as vmstate
Resumed VM, removing state
Logical volume "vm-103-state-suspend-2020-05-10" successfully removed

Pretty neat

 

[Jannoke]

Yes I tested that out also. Got an extra file in my zfs pool. Also it seems it wil dump only needed memory and size of the "hibernate" file depends on the current ram usage. So for faster standby/resume echo 3 >/proc/sys/vm/drop_caches is one way to go. I use it also to make migrations faster between nodes.

 

[davidand]

@Jannoke Interesting approach! Do you issue it on the host machine or inside the VMs? Do you need to wait for a certain time after the 'echo' command?

 

[budy]

Well… the quest could probably hibernate itself after flushing the caches, if one would let it use the PVE api to hibernate itself. Otherwise, the qm command would have to be issued from the PVE host and that one has of course no knowledge as of when or if the cashes have been flushed out.

 

[masgo]

Well… the quest could probably hibernate itself after flushing the caches, if one would let it use the PVE api to hibernate itself. Otherwise, the qm command would have to be issued from the PVE host and that one has of course no knowledge as of when or if the cashes have been flushed out.

Seems to me like a job for the Agent. Would be really nice to have an option where qm suspend would flush the cache from within the guest.

Obviously, this should not be the default.

 

[Jannoke]

@Jannoke Interesting approach! Do you issue it on the host machine or inside the VMs? Do you need to wait for a certain time after the 'echo' command?

The command is blocking. It means it will not return before it's complete. If you have a lot of ram (100+GB etc.) then you can actually see it decreasing from proxmox webui under status. Depending on the machine the flushing can occur on ranges of 5-30GB/s.

 

[Razva]

Yes I tested that out also. Got an extra file in my zfs pool. Also it seems it wil dump only needed memory and size of the "hibernate" file depends on the current ram usage. So for faster standby/resume echo 3 >/proc/sys/vm/drop_caches is one way to go. I use it also to make migrations faster between nodes.

Can you please detail this a bit? Thank you

 

[Jannoke]

I can't much as I don't understand what information you are seeking on this matter
https://www.tecmint.com/clear-ram-memory-cache-buffer-and-swap-space-on-linux/