1

Good evening Server Fault!

I was recently contracted to help a company clean up and consolidate their network\server equipment, and I'm kind of stuck on one issue. I set them up with a new r1soft box for backups. This server is located on the same network as the 2 servers which are being backed up to it. Although the 3 servers are on different VLANs, they are all plugged into the same Cisco Catalyst gigabit switch.

I'm still performing the original backup, however it is going at less than one megabit per second for each server. Occasionally, I will see the speed drop to 0, and then climb all the way back up to a whopping 800Kbps.

Here's a quick overview of the equipment involved:

Backup Server:

  • CentOS 6.7, latest kernel, up to date
  • Server Backup Manager SE, latest version (installed today)
  • SELinux disabled
  • Physical

Servers Being Backed Up: (2 with identical configs)

  • CentOS 6.7, CloudLinux
  • Server Backup Agent, latest version (updated today)
  • SELinux disabled
  • Xen HVM

The 2 servers were originally being backed up to an off-site server. I took a look at the Task History on the original backup server, and I see that the average throughput for these jobs was as low as 499.5 KB/s, and no higher than 4MBPs.

I tested the speed between these 2 servers using 2 methods. First and foremost, an iperf test showed 930MBPs between the 2 servers. Also, I created a 1GB test file, put it on the web server, and curl'd it from the backup server. I got about 120 MBPs. Although these are quite different, they're still much better than anything I've seen on R1 between these 2 machines.

So here's my question- I have a hunch that something was put in place to limit the bandwidth in order to prevent overages with their colocation provider. There doesn't appear to be any traffic shaping on any of the servers involved.

Can someone please recommend a best next step? Thanks!

  • First, disable offloading on the vif interfaces on the hypervisor. Tso offloading that is. – John Keates Jan 18 '16 at 4:59
  • @JohnKeates I used ethtool to disable tso offloading. Didn't really help too much. Still fluctuating around 1Mbps – Zack Magee Jan 18 '16 at 5:43
1

Well, the network of the server of which you are trying to backup is decent, so it might just be R1.

Where is the server located? From what I can see, R1 seems to be in Denver, Colorado (198.64.248.226). The backup speeds won't be amazing if the server is located on the other side of the world, so until that information is provided, your best bet right now is to find a server nearby to the provider you're backing up. You can then send the backup there (closer networks, better peering = faster transfer speeds).

  • Hey - both of the servers are in the same datacenter. I would understand and expect these speeds if I was backing up in 2 distant datacenters, but not when they're connected to the same switch. After doing some reading, it appears you may be right, it might be R1- I see a lot of other complaints about backup speed. It would be nice if there was any way to speed it up.. – Zack Magee Jan 18 '16 at 3:51
  • yeah definitely was just r1soft. Replaced the whole setup with Bacula and I'm getting an average of 840Mbps - and it's free. – Zack Magee Jan 20 '16 at 3:59

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.