CIFS shares causing NAS out of memory errors (consuming over 8 GB RAM)

[autumnwalker]

Hello,

I have several Linux boxes connecting to my NAS (powered by Unraid) via CIFS / SMB. Proxmox, in particular, is causing the NAS to run out of memory using over 8 - 10 GB of RAM over a period of days. This causes the NAS to trigger out of memory errors and the reaper to kill processes (smbd) and results in "stale file handles" across all the boxes which are connected (including Proxmox).

Any idea why this is occurring? Only Proxmox creates the memory issue on the NAS.

 

[wolfgang]

Hi,

This sounds like cache problems on the Unraid Box.
The point is you load an image and not small files as you do with Windows file share.

 

[autumnwalker]

Thanks @wolfgang. I have a thread open on the Unraid forums as well - I'll update that thread with this insight.

Is there a better way for me to connect Proxmox to Unraid in this case? I read an old thread about disabling caching via Proxmox, but it looks like a dated solution.

 

[wolfgang]

I don't know Unraid in detail but try it with NFS.

 

[autumnwalker]

Thanks again! I tried NFS before and didn't have much luck. I can try it again though if that is the preferred method over CIFS.

Is there any way to force CIFS to not cache on the Proxmox side? I know it's a configuration option when you're using regular CIFS mounts (i.e. from a vanilla Linux box), but unclear if it is supported / advisable with Proxmox.

 

[wolfgang]

You can't control the cache of the Server from the Client side.
What you mean is the Client cache, but this has no effect on the server-side.
The problem is that Proxmox VE opens large files on the server-side and make random reads/writes in it.