How is CPU usage in the web GUI calculated? I am getting wildly different CPU usage answers in the GUI, htop, and the proxmox host. See in the below screenshot. Does this mean anything?

 

I am trying to figure out if I have an issue here. I have a share I think is being slow to respond, and causing smb operations to hang on clients or give really inconsistent performance. Could be a drives, they are new to this setup. If its a CPU problem I can allocate more CPUs, but I am not sure thats the issue.

The CPU utilization doesn't make sense to me. It isn't always at 100%. Its constantly dancing around (often hitting 100%) but the GUI always reads way higher than any other method.

 

23 minutes ago, nothet said:

The CPU utilization doesn't make sense to me. It isn't always at 100%. Its constantly dancing around (often hitting 100%) but the GUI always reads way higher than any other method.

The GUI also shows I/O wait and if you have probably issues with a share that could cause the problem.

 

You can see a similar behaviour if you for example install LANCache on a HDD because it causes high I/O wait when running it on a HDD an a client is pulling data from it.

Oh! Ok that makes sense that it includes IO Wait. I have a bunch of spinning HDDs in a SAS DAS. I've got a mix of IO (Jellyfin read and file writes in).

Should I probably look into an SSD cache drive?

Thanks for the quick reply, I appreciate it.

8 minutes ago, nothet said:

Ok that makes sense that it includes IO Wait

At least that‘s my assumption, as said above the same happens when using LANCache when a clien pulls data from a HDD because it have to wait for the HDD because it‘s to slow to pull the requested chunks from the HDD.

 

I assume a SSD for the pool would solve your issue.