22.05-RC high CPU load

[renegade]

After upgrading to 22.05 my cpu is constantly above 1.00 load and gets very warm. Yesterday evening it was above 2.00 and got so hot that the CPU turnend off.
In 22.01 load was 0.05 - 0.10.
On ssh the highest process is php_pfb with around 5%.

Can I switch back to 22.01 with the new boot environment feature?

[MoonKnight]

@renegade
Hi I don't see that problem here after upgrading to 22.05 RC.

[renegade]

Thanks for the feedback.

I have upgraded on Thursday.
This is how it look like.

There is a high increase on CPU.

[MoonKnight]

@renegade
hmm interesting, did you reboot your firewall after the updating to 22.05 RC?
What packages have you installed?

[renegade]

@moonknight
pfblocker. And it seems to be the same issue as described in 3-1-0-4-high-cpu-load
I killed the php_pfb task and cpu load dropped down to 0.02 as before.

I had no issues with pfblocker in 22.05

[MoonKnight]

@renegade
Did you try to reinstall pfBlockerNG-Dev?

I'm running pfBlockerNG-Dev, never seen as high CPU usage as you have.
And I'm running other packages also.

[renegade]

@moonknight
Yes, I did and no change unfortunately.
I now deactivated the services.
When the final arrives, I will make a clean install I guess.

[stephenw10]

@renegade said in 22.05-RC high CPU load:

got so hot that the CPU turnend off

That should not be possible. You should be able to run the CPU 100% without running into thermal shutdown. If you can't then I suggest checking and/or improving your cooling solution.

I have one system here running pfBlocker that behaves like that. After some minutes (not immediately) one CPU core get stuck at 100% until pfBlocker is disabled. Is that what you're seeing?

Steve

[renegade]

@stephenw10
Yeah, you are right. I have to improve that. Unfortunately the device is in a quite warm environment.

And yes, that's what I can see on my installation. Same behaviour.
thanks for confirming, Steven.

[dennypage]

FWIW, I've seen similar.

Hasn't resulted in a problem though.

[stephenw10]

Also running pfBlocker? Does it start after some time, not immediately?

[dennypage]

@stephenw10 Yes, running pfBlockerNG-devel.

At first I thought I noticed that it didn't start right after boot, but I didn't investigate sufficiently.

I just rebooted and checked carefully, and you are right, CPU starts ramping up approximately 2 minutes after boot is complete.

[dennypage]

@stephenw10 Without rebooting, stopping the pfb_filter returns the cpu to normal levels.

Restarting pfb_filter cause cpu to return to higher levels after approximately 2 minutes and 30 seconds..

[renegade]

I noticed that there is a new Snapshot version of the 22.05 RC this morning.
Is it fixed already? Or has pfblocker maybe to fix it?

[stephenw10]

It is not fixed as far as I know. Tracking it here: https://redmine.pfsense.org/issues/13154

Add any additional details you may uncover there.

Steve

[dennypage]

@stephenw10, I've updated the redmine issue with a temporary fix. I make no representation that this fix is the best approach.

@BBcan177, please take a look. Thanks.

[stephenw10]

Nice. Looks like work is already in progress here: https://redmine.pfsense.org/issues/13156

[dennypage]

@stephenw10 Well shucks. Seeing that would have saved me a bunch of work. I guess I need to spend more time spelunking redmine.

[stephenw10]

Yup I missed that too. Or at least I failed to realise the root cause was the same.

Still when you independently come up with the same fix that's always a good sign.

Working for me here on 2.7.

Steve

[drewsaur]

Thank you all and @BBcan177 for the simple patch, because I was about to make a long post looking for help, and the little edit to line 4139 solved my CPU load issues with 22.05 completely.

I think this fix should be made more prominent - I am not using a 22.05 development snapshot, but rather 22.05 final, and I found this thread via a search and not via a browse of the forums. I feel lucky to have found it.