Slow to backup NFS shares

[edumartinDC]

Good morning.

I have a consult to see if you can help me.

I have the following environment set up:

VM Windows server 2016 with Veeam Backup and Replication 9.5 update 4 (8 vCPu, 16 Gb RAM)
Two physical proxies with Windows sever 2019 and direct connection to a NetAPP (2 sockets, 12 cores, 128 Gb RAM)
Two VM CentOS 8 with NFS mount points and Veeam backup linux client

I have two jobs separated by proxy from the NFS mount points of each linux client
each job is 5.22 Gb and 5.90 Gb respectively

The backup is going very very slow for me... process rates of 2 MB/s, the rest of the VM backup jobs are working perfectly, I just have
slow in these two backups that make copy through the linux client of the mounting points.

Any idea where the problem might be?

Greetings

[PetrM]

Hi Eduardo,

Where is the "bottleneck" according to job statistics and what type of repository do you use?

Thanks!

[edumartinDC]

Good morning.

the thing is that the job is still running, I don't know if I could see it somewhere on the client or the server
that bottleneck information

On the other hand, the repositories are Windows computers (the same as the proxy), connected to the NetApp cabinet as a network unit

I don't know if this information is enough or if you need more information


thank you for your time

[veremin]

You can open a backup session of currently running job and the bottleneck component will be provided there among other metrics. To open a backup session you can simply double-click on a job in the backup server UI. Thanks!

[edumartinDC]

Hello.

this is a message that appears in the log of the linux agent where the job is running

Code: Select all

 pex: 2446698987526/2446698987526/0/2446698987526/1442491841211 / 4/0/1/0/7/124
[21.04.2020 15:34:01.729] <139808944932608> vmb | Session progress: 74%; processed: [2446698987526/257015087104] read: [2446698987526], transfer: [1442491841211] speed: 1559949 bottleneck: 4/0/0/124

Greetings

[edumartinDC]

Ok, I saw what you said, I have the bottleneck in the target, I tend to be the repositories, so it may be a problem of the cabin or communication to it.

I'm going to check this point... and I'm going to report

Greetings and thanks

[veremin]

Just curious - what is used as a target repository? How well it can handle random I/O?

[PetrM]

By the way this is a good question... As far as I understand:

the repositories are Windows computers (the same as the proxy), connected to the NetApp cabinet as a network unit

This tool can simulate different I/O load on repository depending on backup chain, it's worth trying to estimate disk performance.
This page of our best practices reflects I/O impact depending on different backup modes.

Thanks!

[edumartinDC]

Good morning.

Well, at the moment I have formatted the shared drive in the windows computer with NTFS instead of ReFS and the destination is no longer a bottleneck, now I have average transfer rates of 19 MB/s, I will continue to observe because now he tells me that the bottleneck is at the source, but that already fits me a little more since the server where I have the NFS mount points is a VM with a linux and the folders have many files ... probably small.

I'm telling you how it is... thanks for your contributions.

Greetings