-
- Veteran
- Posts: 282
- Liked: 25 times
- Joined: May 22, 2015 7:16 am
- Full Name: Paul
- Contact:
Copy & Tape jobs slow also copying to network shares slow
Had problems with copy and tape performance for some time and also copying files to network shares from the Veeam Backup & Replication server. Copying to the same locations from a proxy was OK. Backup performance was not affected.
I had opened a support call but the point about performance of the backup server was not picked up and the support call was closed with the root cause being the storage. I have been investigating this (obviously) and found some documentation regarding VMXNET3 performance issues. I was planning to remove the adapter and add it again but did the following instead only on the backup server.
I ran netsh int tcp show global and everything looked OK. RSS was enabled
I ran netsh int tcp set global RSS=Enable
Obviously the result said it was enabled as it was.
What has happened since though is that the send and receive on the NIC is now peaking at 6 GB where before the most was 2 GB and copy jobs that had been failing have now completed. I have no idea why this would have changed anything as RSS was already showing as enabled.
I am only adding this to the forum in case anyone else is experiencing similar issues.
Repositories are all CIFS
Still confused why so much traffic goes through the backup server and not the proxies though. The backup server is not configured as a proxy.
I had opened a support call but the point about performance of the backup server was not picked up and the support call was closed with the root cause being the storage. I have been investigating this (obviously) and found some documentation regarding VMXNET3 performance issues. I was planning to remove the adapter and add it again but did the following instead only on the backup server.
I ran netsh int tcp show global and everything looked OK. RSS was enabled
I ran netsh int tcp set global RSS=Enable
Obviously the result said it was enabled as it was.
What has happened since though is that the send and receive on the NIC is now peaking at 6 GB where before the most was 2 GB and copy jobs that had been failing have now completed. I have no idea why this would have changed anything as RSS was already showing as enabled.
I am only adding this to the forum in case anyone else is experiencing similar issues.
Repositories are all CIFS
Still confused why so much traffic goes through the backup server and not the proxies though. The backup server is not configured as a proxy.
-
- Product Manager
- Posts: 6551
- Liked: 765 times
- Joined: May 19, 2015 1:46 pm
- Contact:
Re: Copy & Tape jobs slow also copying to network shares slow
Hi,
Thanks!
Does the proxy have the same network parameters as VBR?Copying to the same locations from a proxy was OK
Do you expect the traffic to go through the proxies when you copy files from VBR? Which server holds the role of tape server? Where do you have you shares mounted?Still confused why so much traffic goes through the backup server and not the proxies though. The backup server is not configured as a proxy.
Thanks!
-
- Veteran
- Posts: 282
- Liked: 25 times
- Joined: May 22, 2015 7:16 am
- Full Name: Paul
- Contact:
Re: Copy & Tape jobs slow also copying to network shares slow
Proxy servers are VMs in the same vSphere environment and have the same network configuration as the backup server
I don't expect files to be copied from the proxies if I am directly copying the files through explorer on the backup server however I thought the point of the proxy was to transport data to the repositories? As I said in the post the performance was only impacting copy and tape jobs. The tape jobs use the copy files as their source
The tape server is a separate physical server
I don't expect files to be copied from the proxies if I am directly copying the files through explorer on the backup server however I thought the point of the proxy was to transport data to the repositories? As I said in the post the performance was only impacting copy and tape jobs. The tape jobs use the copy files as their source
The tape server is a separate physical server
-
- Product Manager
- Posts: 6551
- Liked: 765 times
- Joined: May 19, 2015 1:46 pm
- Contact:
Re: Copy & Tape jobs slow also copying to network shares slow
Proxies are not utilized during file-copy jobs.however I thought the point of the proxy was to transport data to the repositories?
File To Tape jobs move data between target and source datamovers. Target datamover sits on a Tape server, source datamover resides on the server where you copy your data from. In your case, if source files are located on CIFS repository, then the source datamover resides on the gateway for the repository.
Thanks!
-
- Veteran
- Posts: 282
- Liked: 25 times
- Joined: May 22, 2015 7:16 am
- Full Name: Paul
- Contact:
Re: Copy & Tape jobs slow also copying to network shares slow
I am not doing file copies and the tape jobs are using copy jobs as their source not a repository. I still do not understand why backup jobs work great but the copy jobs are slow and appear to add complexity. Why do copy jobs not just copy the backup files without adding any additional processing?
The performance of the backup server improving after I ran the RSS command is still very strange
The performance of the backup server improving after I ran the RSS command is still very strange
-
- Product Manager
- Posts: 6551
- Liked: 765 times
- Joined: May 19, 2015 1:46 pm
- Contact:
Re: Copy & Tape jobs slow also copying to network shares slow
Sorry, it wasn't quite clear what copy jobs you were referring to. Backup copy jobs do not use proxies either - data is copied directly between datamovers residing on repositories (unless you have WAN accelerators deployed). You use CIFS repositories, so datamovers will reside on gateway servers. That is why it is always recommended to keep a gateway server as close to its associated CIFS storage as possible.
Regarding RSS...although that might be a needless question to ask, but I wonder if you had had a chance to check in the Task Manager if only one CPU core was being 100% utilized on the gateways before you tried to toggle RSS on? I am asking because what you've described in your first post sounds like a minor OS glitch - Windows reported the function to be already turned on, while in fact it wasn't.
Thanks!
Regarding RSS...although that might be a needless question to ask, but I wonder if you had had a chance to check in the Task Manager if only one CPU core was being 100% utilized on the gateways before you tried to toggle RSS on? I am asking because what you've described in your first post sounds like a minor OS glitch - Windows reported the function to be already turned on, while in fact it wasn't.
Thanks!
Who is online
Users browsing this forum: No registered users and 28 guests