Feature request: Backup Scheduler

J

JCANINM

New Member
Proxmox Subscriber
Nov 14, 2025
4
4
3
I have an 8-node setup and run daily backups of (almost) all my vm's. Initially i just had a single job with all vm's listed, starting at 22:00, every day.
I soon found out that that didn't work for me. The datastore on the underlying PBS is a simple raid with spinning sas disks, 80Tb but very slow, and my backup job fired up 8 backup simultaneouly (one from each node) , chooking the poor raid.
So now i have a job for the vm's on each node. Each job has an aproximate start time of the "calculated length" of the previous job.
Now to the point: The schedule is missing an important option. Instead of giving the job a start time, it would be nice to also have an option to start the job after a specific job has finished, avoiding idle time between jobs or overlapping jobs.
The result would be, for example:
Job1 starts at 22:00
Job2 starts after Job1 is done
Job3 starts after Job2 is done
etc...

I have tried to search the forum but havent found what I'm looking for - so please forgive me if this has already been asked about before.
 
  • Like
Reactions: SteveITS
There is a feature request on bugzilla to tha same effect: https://bugzilla.proxmox.com/show_bug.cgi?id=6647

What I propose is a timer for jobs that can be interrupted as to allow other jobs to start at a certain time - plus a setting to start a job only if no more than N other tasks are already running.

The proposal is for the PBS side. Essentially, it is the target that is overwhelmed, especially if many sources use the same backup target.
 
Last edited:
  • Like
Reactions: JCANINM
JCANINM said:
PBS is a simple raid with spinning sas disks, 80Tb but very slow, and my backup job fired up 8 backup simultaneouly (one from each node) , chooking the poor raid.
Click to expand...
How many backup jobs are running at the same time is (nearly) irrelevant. PBS needs to access (pseudo-) random .chunks in any case. For writing new, modified data, that is.

For already existing chunks it just reads only metadata to know this very fact. That's one reason why all of my PBS' have a fast Special Device. Mirrored of course.
 
  • Like
Reactions: JCANINM
The problem also arises with sync and verify jobs, where parallelisation on spinning disks slows things down quite a bit. Alas, there can also be ZFS maintenance jobs like zfs scrub which are not handled via Proxmox tasks at all. Those job types are a good example of being restartable at almost any point (for backup jobs, you have to complete at least a single machine at a time).
 
  • Like
Reactions: JCANINM
SteveITS said:
Another point of view for this would be, instead of one job per node, one job with the nodes listed in order.
Click to expand...
I have tried this. Whenever a backup job includes multiple nodes, each node will start baking up it's vm's at the same time, resulting in multiple backups hitting PBS at the same time. If you know a way to only have one vm being backed up at a time, pleas elaborate as that is exatly what im looking for.
 
meyergru said:
There is a feature request on bugzilla to tha same effect: https://bugzilla.proxmox.com/show_bug.cgi?id=6647

What I propose is a timer for jobs that can be interrupted as to allow other jobs to start at a certain time - plus a setting to start a job only if no more than N other tasks are already running.

The proposal is for the PBS side. Essentially, it is the target that is overwhelmed, especially if many sources use the same backup target.
Click to expand...
Not exactly what i'm looking for. Right now i dont care about PBS, I want to control how many simultaneous jobs are sent to it. The backup could also be on a shared drive or SAN. I beleive the scheduling is a Proxmox VE issue, and not PBS :) Hence the request for some kind of "chaining" backup jobs.
 
  • Like
Reactions: Johannes S
UdoB said:
How many backup jobs are running at the same time is (nearly) irrelevant. PBS needs to access (pseudo-) random .chunks in any case. For writing new, modified data, that is.

For already existing chunks it just reads only metadata to know this very fact. That's one reason why all of my PBS' have a fast Special Device. Mirrored of course.
Click to expand...
Could be, but i dont care about PBS in this scenario. I want the ability to control how many jobs are pushed to PBS simultaneously. Hence the request for being able to "chaining" backup jobs. The scheduler for this, i believe, lies in Proxmox VE :)
 
  • Like
Reactions: Johannes S and UdoB
You must log in or register to reply here.
Top Bottom