How do I determine the cause of high CPU utilization?

[FreeMan]

My CPU has been like this for the past couple of hours and it will peak like this every now and then. I have no idea what causes this.

I've looked at docker usage and they're usually 1% or less (I currently have 2 dockers running around 15% each, but that still only accounts for 30% total, not the 100% being shown.)

 

This is one of the very few times top has shown my anything running above 10% or so, so this does seem almost reasonable:

the tdarr_server I'd expect, since I'm doing a lot of file transcodes (but none are actually transcoding on the UNRAID server itself - all nodes are running on Win machines).

How do I determine what the smbd processes are that are at 21% and 11%?

Obviously shfs was running very high, how do I determine what was causing that?

 

CPU utilization has just dropped off, but this is the first time all afternoon it's been at a reasonable level:

 

Again, how do I go about figuring out what causes this high CPU?

 

• Is there a way to log from top?
• Is there something else that'll show and log what's using CPU?
• How do I determine what's causing SHFS and SMBD to use so much CPU?

 

 

Under the assumption that it's usually helpful: nas-diagnostics-20240308-1526.zip These were taken after CPU utilization dropped back down, though.

[FreeMan]

Just spiked again. This time, I know that my Sonarr docker is renaming a lot of files to fix up their naming, but, how do I determine what is actually happening when I haven't just kicked off some sort of process that I know may cause high CPU.

 

 

during the CPU spike: nas-diagnostics-20240308-1534.zip

[MAM59]

You need to cool down :-)))

It is a known problem of UNRAID to show IO load as 100% CPU load.

If you copy a lot of files and your drives are not fast enough, buffers run full and transmission is halted. Although the CPU really is idle waiting for a free disk slot, utilization is shown as 100%.

If you want to know if this happens to you or if something else hogs the CPU, take a look at the CPU temperature. If it is rather low, its the false IO reading, if it is glowing in the dark, something else is going on.

 

Remember the Hitchhiker: DON'T PANIC! 😁

 

 

Edited March 9 by MAM59

[FreeMan]

Fair enough, @MAM59. I've had ongoing issues with this over the last couple of years and have never managed to figure out the cause. I don't have the fastest server, MoBo, drives or controllers, so this isn't a totally unreasonable suggestion.

 

If you'll note in the screen shots, though, CPU temp was 51 & 52°C when it was reporting 100% CPU and had dropped to 39°C at 77% total CPU load, so I'd say the CPU was actually busy doing something, not just waiting for I/O.

 

Also, the web interface of several dockers was running slowly, as though the CPU was highly stressed and my requests to the various dockers were in queue waiting for some CPU attention.

 

If you've got any other suggestions on how to determine what's actually going on, I'm more than open to listening.

[MAM59]

1 hour ago, FreeMan said:

If you'll note in the screen shots, though, CPU temp was 51 & 52°C when it was reporting 100% CPU and had dropped to 39°C at 77% total CPU load, so I'd say the CPU was actually busy doing something, not just waiting for I/O.

I would not say 52° was anything near "hot" or stressed. Up to 72° are allowed for this type of CPU.

39° even seems to be idle state.

 

I just run an intensive memorytest of the new PC I'm just building, its an AMD 7950x (16 Core) and stays at 63° even with all cores at 100% (and slightly overclocked ram).

I guess I will run it a few hundreds Mhz faster, up to 95° are allowed for it (but to save my ears I will only allow it up to 80° I think)

 

[FreeMan]

52°C is at the upper end of the range I've ever _noticed_. Doesn't mean it hasn't been hotter, that's just the hottest I've seen...