Why backup needs a global lock?

[kernelpanicattack]

When one backup is running, the second backup of another VM cannot start with

INFO: trying to get global lock - waiting...
ERROR: can't acquire lock '/var/run/vzdump.lock' - got timeout

This has been discussed million times on this forum, but the answers generally are either do not start two backups at the same time or increase the timeout so the second job can start later. All these threads miss one important thing -- why does PVE needs a global lock at all? Why can't different VMs be backed up at the same time?

The ability to run backups in parallel is very important for many in many use cases. In our case, for example, we have a scheduled job to backup all VMs that runs all night, and customers schedule their own backup jobs (that even go to a different backup server!). The customers jobs do not start in this case.

 

[fiona]

Hi,
the rationale is that each backup can produce a lot of load, so it was limited to one backup at a time on a given node. There is a feature request to make the number configurable: https://bugzilla.proxmox.com/show_bug.cgi?id=3347

 

[stv.Bundesadmin]

Hello @fiona, could this feature request please please please be worked on?
Running backups of fileservers > 10TB (e.g. after resizing, no dirty bitmap to use etc) locks out regular "small VM Backups" for many hours.
We're running all-nvme and have flash-based backup-systems as per Proxmox-recommendation.
Modern hardware all around, far from being performance-restricted.

We could run a lot of backups in parallel to achieve our intended RTO/RPO, yet this limit of "only 1 backup per node" artificially limits us in a bad way.

 

[fiona]

Hi,

Hello @fiona, could this feature request please please please be worked on?

unfortunately, I don't have time to work on this currently and am also not in charge of assigning such work.