Restore VM seems to kill other running VMs

[Jim J]

I am in the midst of migrating from Prox 3 to 4 and have run into an issue while restoring VMs from the old to the new machine.

I have restored a few VMs onto my new 4x machine, but each time I restore another machine, the restore action seems to kill my running VMs while the restore is running. They become unresponsive until after the restore job is complete, then all resume normal operation.

Here are some storage results in case they are useful:

root@prox1:~# pvesm lvmthinscan pve
data
root@prox1:~# pvesm lvmscan
pve
root@prox1:~# pvs
  PV  VG  Fmt  Attr PSize PFree
  /dev/sda3  pve  lvm2 a--  1.36t 15.84g
root@prox1:~# lvs
  LV            VG   Attr       LSize   Pool Origin Data%  Meta%  Move Log Cpy%Sync Convert
  data          pve  twi-aotz--   1.24t             10.37  5.40                           
  root          pve  -wi-ao----  96.00g                                                   
  swap          pve  -wi-ao----   8.00g                                                   
  vm-102-disk-1 pve  Vwi-a-tz-- 100.00g data        0.43                                   
  vm-104-disk-1 pve  Vwi-aotz--   8.00g data        80.81                                 
  vm-105-disk-1 pve  Vwi-a-tz--  60.00g data        70.05                                 
  vm-106-disk-1 pve  Vwi-a-tz--  50.00g data        98.73                                 
  vm-106-disk-2 pve  Vwi-a-tz--   1.00g data        1.27                                   
  vm-110-disk-1 pve  Vwi-aotz--  34.00g data        99.65

 

[sahostking]

Hi

We noticed the same thing occur on our side. But we could never really figure it out. I think its got something to do with lvm thin? As that is possibly the only change from proxmox 3? Maybe someone else could clarify that part.

 

[wolfgang]

Hi,

this is "normal" because the restore write as fast it goes and this will affect the running VM.

 

[fireon]

Hi,
this is "normal" because the restore write as fast it goes and this will affect the running VM.

Really, normal? Never seen this before. So actually when i restore some VM, some other VMs can crash?? Is this depending on I/O, HDD speed SAS /SSD /SATA?

Thanks

 

[wolfgang]

As Jim J wrote

They become unresponsive until after the restore job is complete, then all resume normal operation.

so no VM really crash, it come unresponsive because it waits for io.

It dependence on the setup and the load.
If you backup disk is a slow usb2 disk and you restore to a ssd you will not see any affect, because the backup disk is slow and the restore target disk is fast and can handle this io's.

But if you restore form two disk which are the same class and on the restore target disk run more VM you have an io bottleneck.

 

[fireon]

Ah, ok Thank you.

 

[sahostking]

Anyway to limit restore bandwidth like backups can be done? For backups we use /etc/vzdump.conf and set bandwidth limit in there is there any possible way to do the same for restores?

 

[wolfgang]

At the moment it is not possible to restore with a speed limit but its on the list.

 

[fabian]

if you really need it, you can (as root on the CLI) pipe your backup archive to qmrestore and put some rate limiting pipe inbetween (cstream, pv, ..). but yes, rate limiting for various I/O intensive tasks is on the agenda!

 

[sahostking]

Thanks for this