Reduce amount of Proxmox-Backup-proxy workers for better write speed

[Ivar418]

Hello,

I am using pbs on a very basic system. Just a nuc with a big hard drive attached to it. (wd My Book 18TB)

I noticed that write speeds are very low, around 30MB/s eventho this could be much higher. What I see is that there are many proccesses writing tot the drive at once. Since this is a hard drive it does not like that many simultanious opperations.



My question is, is there a way to reduce the amount of proxmox-backup-proxy workers to get better write speeds? The benchmark shows that the performance could be much better,


I found something regarding editing /etc/vzdump and then the amount of workers, but I couldnt figure it out (i changed it to 4 but still many more workrers spawned) so i assumed that was amount of workers on the pve side and not for the pbs side.


PVE that is being backed up contains 2 raidz1 vdevs and the read speeds is be much more than 30MB/s and shouldnt be a bottleneck.

 

[Dunuin]

Yes, that is only for PVE as most stuff (encryption/compression/hashing) is done on the client side.

Just a single 18Tb HDD is terrible for a PBS. PBS needs IOPS performance. Highly recommended (PBS hardware recommendation also tells you this) to use a lot of small striped HDDs with a SSD for metadata or even better a SSD-only storage.

 

[Ivar418]

Yes, that is only for PVE as most stuff (encryption/compression/hashing) is done on the client side.

Just a single 18Tb HDD is terrible for a PBS. PBS needs IOPS performance. Highly recommended (PBS hardware recommendation also tells you this) to use a lot of small striped HDDs with a SSD for metadata or even better a SSD-only storage.

So no way to reduce the amount of proxmox-backup-proxy workers?