I have noticed something strange in the transfer speed when backing up my VMs. I have PBS running in a VM and mounted an NFS share from my Unas Pro. When I back up to PBS, the transfer speeds are around 300 MiB/sec, but when reaching 50%, the speed decreases to 15-30 MiB/sec. What could be causing this issue?
Speed decreases when doing a backup
- Thread starter HomeLabNerd
- Start date
Thank you for your answer.
I'm trying to figure out to run it in a VM or build something for my rack with its own storage. So you advise local storage with zfs. Can you explain, is the ssd a sort of caching for the HDD zfs mirror? I don't have much experience with zfs.
I'm trying to figure out to run it in a VM or build something for my rack with its own storage. So you advise local storage with zfs. Can you explain, is the ssd a sort of caching for the HDD zfs mirror? I don't have much experience with zfs.
Last edited:
Only VM data is read, if data already exist on PBS no data is sent/written to PBS.
new/changed data is sent/written to PBS, often limited by HDD PBS storage because data is written in many small chunks instead sequential. (backup data as chunks allow dedup and forever incremental backups)
I have a link for you: https://pbs.proxmox.com/docs/installation.html#system-requirements
So the small chunks is explaining the slow write speeds. It does a sort of caching, till it reaches 50%.
Thank you for your answer!
Last edited:
My homelab friend, this sort of behavior is relatively common out of any sort of system that transfers large files over the network.
There will be multiple levels of cache along the way, some of which you control, and others you may never become aware of.
You are largely confronting network speed and disk speed. In both of those areas you will encounter cache that makes it look like quite a bit more has happened than has really taken place.
Just for a point of reference, I recently made some network changes on a Veeam/VMware system, and was gloating over the incredible speeds of my first tests ... till I started testing with VMs over about 50gb, and then I saw the same old crawl I was used to.
I don't mean to discount the things folks have told you here. They are dead-on.
If you want your backups to perform well, you will reconsider your NAS mount.
There's more at play here than just the NAS being slow. PBS performs really poorly on an NFS/SMB mount.
The subject of PBS and a NAS comes up a lot. Do some reading.
Um ... this is something i wrote about PBS and NAS. Maybe start here.
There will be multiple levels of cache along the way, some of which you control, and others you may never become aware of.
You are largely confronting network speed and disk speed. In both of those areas you will encounter cache that makes it look like quite a bit more has happened than has really taken place.
Just for a point of reference, I recently made some network changes on a Veeam/VMware system, and was gloating over the incredible speeds of my first tests ... till I started testing with VMs over about 50gb, and then I saw the same old crawl I was used to.
I don't mean to discount the things folks have told you here. They are dead-on.
If you want your backups to perform well, you will reconsider your NAS mount.
There's more at play here than just the NAS being slow. PBS performs really poorly on an NFS/SMB mount.
The subject of PBS and a NAS comes up a lot. Do some reading.
Um ... this is something i wrote about PBS and NAS. Maybe start here.
Last edited:
Thank you for your answer. Checked the link!My homelab friend, this sort of behavior is relatively common out of any sort of system that transfers large files over the network.
There will be multiple levels of cache along the way, some of which you control, and others you may never become aware of.
You are largely confronting network speed and disk speed. In both of those areas you will encounter cache that makes it look like quite a bit more has happened than has really taken place.
Just for a point of reference, I recently made some network changes on a Veeam/VMware system, and was gloating over the incredible speeds of my first tests ... till I started testing with VMs over about 50gb, and then I saw the same old crawl I was used to.
I don't mean to discount the things folks have told you here. They are dead-on.
If you want your backups to perform well, you will reconsider your NAS mount.
There's more at play here than just the NAS being slow. PBS performs really poorly on an NFS/SMB mount.
The subject of PBS and a NAS comes up a lot. Do some reading.
Um ... this is something i wrote about PBS and NAS. Maybe start here.
I'm also testing now with a real PBS server to see the differences.
Btw is it possible to control the amount of VM's it copies at the same time? I have a 3 node cluster and it copies a VM from each node simultaneously. For now the PBS server only has a gigabit port.
If you ran one job with all vms on all servers,
then it would run one backup from each server simultaneously.
So for three servers, yes 3 backups at once. This is normal.
If you want to restrict it, make 3 backup jobs.
Constrain job each to a server. Backup everything there.
Stagger the three jobs.
then it would run one backup from each server simultaneously.
So for three servers, yes 3 backups at once. This is normal.
If you want to restrict it, make 3 backup jobs.
Constrain job each to a server. Backup everything there.
Stagger the three jobs.