Linux Server for accessing NFS shares

[timmi2704]

Hi together,

as it is highly recommended to use NFS-shares instead of CIFS shares, I installed a CentOS server in order to access NFS-shares of our deduplication appliance. Backup jobs using the newly created repository work in general beside some problems (Client errors) I created a support case for (#00629855).

I have some questions regarding this linux repository server.
1. Are there any recommendations for CPU or RAM amount of that linux server VM? I have 4 vCPUs and 4GB of RAM. But the VM is heavily utilized while backups are running (only 1 or 2 jobs at the moment - planned are far far more). Transfer speeds to the repository could be a lot better (~100 MB/s). After I increased RAM size to 8 GB it is still completely in use.

2. I want to access the NFS share for some backup copy jobs running on remote locations via a WAN connection. How will Veeam handle this? For example how will the traffic flow, when the backup copy job is running / when a transformation is running / when a verification is running? When I used the CIFS shares I was able to set a Windows Server at the repository location as a proxy which would handle all of this. Will this now be done by the linux server?

Thank you very much for helping me out, have a nice day.

[Dima P.]

Hello Timo,

System Requirements for Veeam repository is to have at least 4 gigs of RAM. Here is a good blogpost: configuring CentOS NFS Linux repository, which might be useful, there, for example, 4 vCPU & 8 GB RAM config was used to keep 4 concurrent tasks.

Transfer speeds to the repository could be a lot better (~100 MB/s)

What shows the bottleneck statistics in the backup job?

It should start transport service on the Linux repository, described in more details overhere: Help center - Data Transport Path

Thank you.

[Vitaliy S.]

as it is highly recommended to use NFS-shares instead of CIFS shares

Also can you please tell me where did you get it from as I do not remember such recommendation?

[dellock6]

He has a deduplication appliance, so I think he's referring to the often used recommendation to place a linux server in front of the NFS share exposed by the appliance itself, rather than conneting to it via CIFS. Timo, what appliance are you using? Because latest versions of DataDomain or HP StoreOnce firmwares for example have good performances also when used in CIFS mode...

[timmi2704]

Thanks for your replies together

What shows the bottleneck statistics in the backup job?

I tried some more jobs and the Bottleneck statistic shows the source as a bottleneck. I will do some more testing with different jobs so that I'll get some more reliable results.

Also can you please tell me where did you get it from as I do not remember such recommendation?

Luca is right, I was talking about the linux server which enables me to use an NFS share of my deduplication appliance instead of a CIFS share. Sorry If I wasn't clear. This is a Quantum DXi - No DataDomain or HP StoreOnce.

I'm now struggling with some problems with the backup jobs using the new NFS repository. Unfortunately some VMs in this job randomly stop backing up with error messages like "Client error: An existing connection was forcibly closed by the remote host" or "Client error: Connection timed out Unable to retrieve next block transmission command". But this issue is currently being investigated by the support team. When disabling parallel processing it works a bit better.