Recent content by tomstephens89

Please share.

Just as bad, or worse. Upload image '/dev/mapper/pve-vm--100--disk--0' to 'root@pam@10.226.10.10:8007:pbs-primary' as tomtest.img.fidx tomtest.img: had to backup 62.832 GiB of 80 GiB (compressed 42.404 GiB) in 673.35s tomtest.img: average backup speed: 95.552 MiB/s tomtest.img: backup was done...

Some serious newer CPU's than my Xeon Plat 8268's I see. Thats a huge leap in performance. I have just tested using a host with an Epyc 9354P in it and I get 400MB/s on there. This smells like a massive single thread performance limitation to me.

What CPU's do you have in your servers that are getting close to 1GB/s?

What CPU's are in your PVE hosts? I am able to get multi GB/s read/write on both my PVE hosts and the PBS datastore. I am able to max the bonded 10G network between them with iperf. I am able to to get over 500MB/s using SCP.... But proxmox backups, run 150-200MB/s. Changing sync level to...

Have just set recordsize to 1M on my main datastore. Lucky we rebuilt the PBS yesterday to test ZFS as everything on here so far has been with hardware RAID 10. However, even after setting this record size, there is no difference in backup performance. I can't get past 200MB/s. See iperf is...

We have rebuilt the PBS a couple of times during testing and also used more than one node for testing. The PVE's are all the same. To clarify, Jupiter is now running standalone as a PVE host. pbs-primary is the SSD backed PBS as above. Mercury is one of our production cluster nodes. Network on...

Latest builds all round, the PBS was installed 2 days ago and the PVE hosts a week. Yes I am aware of the sparse backups but when its got actual data to copy, it maxes out around 200MB/s. Frequently less. Benchmarks on the PBS storage with fio are even better. Its a 24 SSD RAID 10 array...

Well thats good to hear, I will keep a close eye.

I have testing using my secondary PBS which is all spinning disks, vs my primary which is all SSD.... Performance is very close, almost the same. There is a MASSIVE CPU limitation here.

Interesting but this only looks to apply to PBS at the datastore level for verify 'jobs'. Rather than the chunk verification checksumming process which happens at backup on the host.

I am also suffering poor backup performance but am running powerful, enterprise hardware with all flash storage in hardware RAID 10... I beleive the 'chunk verification' is the bottleneck. But these are modern Xeon Platinum's in my hosts...

Resurrecting this thread as I have the same problem. Performance of backup and restore is terrible, even with 10G network and all flash source/destination servers. See here: https://forum.proxmox.com/threads/pbs-performance-improvements-enterprise-all-flash.150514/ I am also running a...

Further testing reveals no disk bottleneck at source, no disk bottleneck at PBS, iperf able to saturate network, SCP able to exceed 500MB/s.... Yet backups run an average of 200MB/s and restores are much much slower. Anyone care to shine any light on this? My hardware and network is good but...

Ok then, shouldn't have an issue rejoining then as we havn't manually pinned any keys. To remove the dead node, we simply ran PVECM delnode XXXX and then removed the nodes folder from the cluster filesystem.