Slow backup speed

[izlat]

Hi!

This is backup result for Debian 8 server (via NFS, through Samba is even worse):

Duration: 08:11:15
Processed: 520.8 GB (100%)
Processing rate: 18.1 MB/s - The horror!!!

This is network speed to backup destination:

iperf -c 192.168.110.61
------------------------------------------------------------
Client connecting to 192.168.110.61, TCP port 5001
TCP window size: 4.00 MByte (default)
------------------------------------------------------------
[ 3] local 192.168.110.40 port 56910 connected with 192.168.110.61 port 5001
[ ID] Interval Transfer Bandwidth
[ 3] 0.0-10.0 sec 10.2 GBytes 8.73 Gbits/sec

Windows-servers show the speed of backup to the same destination storage 5-6 times faster (about 100..120MB/s)!
Why?

[PTide]

Hi and welcome to the community.

I'm going to ask you a few questions:

What kind of backup is that - file-level or volume?

Did you configure the repo as local (manually mounted the share) or as a SMB/NFS via UI?

Have you tried to run test backup to a locally attached storage?

What is shown as a bottleneck in job stats?

Thanks

[izlat]

Hi and welcome to the community.

I'm going to ask you a few questions:

What kind of backup is that - file-level or volume?

Did you configure the repo as local (manually mounted the share) or as a SMB/NFS via UI?

Have you tried to run test backup to a locally attached storage?

What is shown as a bottleneck in job stats?

Thanks

1. Volume level backup.
2. Repo - NFS via UI
3. No, this possibility I have not tried. But I'll try.
4. Source:

Code: Select all

Duration:           08:11:15         Processed:          520.8 GB (100%)  
Processing rate:    18.1 MB/s        Read:               520.8 GB         
Bottleneck:         Source           Transferred:        6.7 GB (78x)     

Gosha

[izlat]

Hi!

I just tried a variant with local mounted via nfs repo:

Code: Select all

Processing rate:    22.3 MB/s        Read:               14.1 GB          
Bottleneck:         Source           Transferred:        1.8 

The result is slightly better, but still unacceptable.

Manual copy long file (about 4 GB) from local disk to repo as local mounted nfs - shown about 70MB/s.
I do not understand how to attach a screenshot in my post.

Gosha

[PTide]

Source bottleneck actually means that it is the disk what makes the backup to slow down. Do you have any other activity running concurrently with the backup job? Simple file copying works faster because you it does not involve COW snapshot which generates 3x more IOs (1 read + 2 writes).

You can upload a screenshot on some public picture hosting resource and post a link here.

Thanks

[izlat]

Source bottleneck actually means that it is the disk what makes the backup to slow down. Do you have any other activity running concurrently with the backup job? Simple file copying works faster because you it does not involve COW snapshot which generates 3x more IOs (1 read + 2 writes).

Backup the same VM with the Proxmox-cluster means (without the possibility of an incremental backup) occurs
about three times faster (less 3 hours).

I will look for other tools, since Weeam Agent for Linux makes even an incremental backup in about 9 hours.
This is unacceptable for me.

Gosha

[PTide]

There is a big difference between VM-level backup and in-guest backup. Before you search for other tools, would you kindly contact our support team so they can investigate for possible reasons? Should you decide to do so, please post your case ID.

Thank you