Resolved The average load goes up and up

[cmoraleda]

Server operating system version

Debian 12

Plesk version and microupdate number

Plesk Obsidian v18.0.57_build1800231208.08 os_Debian 12.0

Hi, I just installed a new dedicated server with Plesk and something is puzzling me.

It is a new installation, the domain has no traffic and WordPress is installed.

The average load goes up and up and doesn't seem to stop at any value. I have seen it with a value of 40. I have restarted the server and the same thing happens again, it goes up little by little.

It is also true that everything seems to be going well, I can visit the domain in the browser, I can enter the WordPress admin and everything seems correct and loads quickly.

The values for CPU, Memory, Disk, Networking... are all at a minimum since, as I said before, the domain has no traffic. I only have a monitoring tool connected to detect possible falls (UptimeRobot).

Should I worry about this? Is it a Plesk bug?

 

[Peter Debik]

What's the output of ps aux | sort -nrk 3,3 | head -n 20?

 

[cmoraleda]

Nothing appears when I execute the command, it stays as I show in the screenshot.

 

[cmoraleda]

What I observe is that the increase in average load follows a pattern, every hour the average load increases by one unit.

 

[cmoraleda]

What's the output of ps aux | sort -nrk 3,3 | head -n 20?

The ps command does not work or hangs. I have taken out that data that I attach with top

 

[mow]

The average load goes up and up and doesn't seem to stop at any value. I have seen it with a value of 40. I have restarted the server and the same thing happens again, it goes up little by little.

What load is: The number of tasks that are currently waiting.
Optimal load distribution is to have one task per (virtual) core waiting so no core will be waiting for something to do.
If you have more tasks than that waiting, things are starting to lag.
Now if you have tasks that are dead (waiting on a broken I/O device, zombies waiting on dead child processes ...), those will count towards waiting tasks and throw off your load calculation.

You probably have a zombie outbreak. Check your hardware.

 

[cmoraleda]

What load is: The number of tasks that are currently waiting.
Optimal load distribution is to have one task per (virtual) core waiting so no core will be waiting for something to do.
If you have more tasks than that waiting, things are starting to lag.
Now if you have tasks that are dead (waiting on a broken I/O device, zombies waiting on dead child processes ...), those will count towards waiting tasks and throw off your load calculation.

You probably have a zombie outbreak. Check your hardware.

I have run top -n 5 and it shows this which I pasted below. Curiously, in the first run there is a process (agent360) that is using a CPU usage percentage very similar to what appears in the load average.

 

[Peter Debik]

The problem does not lie in the processes that "obviously" occupy cpu time. The problem here is the enormous number of useless kworker etc. processes on the system. As @mow said: hardware issue? Or some awkward operating system issue?

 

[mow]

Might be the i9-13900 with its different core types not being fully supported by the debian 12 kernel yet.

 

[cmoraleda]

Might be the i9-13900 with its different core types not being fully supported by the debian 12 kernel yet.

Thanks for your answers. Do you advise me to change to Debian 11? Or stay with 12? I'm not sure if this could represent a problem in the medium long term.

 

[mow]

Do you advise me to change to Debian 11? Or stay with 12?

The problem is that the default kernel might be too old for your CPU, so going back to an even older debian won't help. This would rather be a case for trying a kernel backported from debian 13.

 

[cmoraleda]

I finally changed machines to one with an AMD processor AMD Ryzen 7 7700 8-Core and Debian 11.8. I haven't had any problems again.