apache nobody process and cpu high load

[benito]

Hello!

The server is really experiencing high load when the stuck httpd processes are more than 6, this happens in a couple of days.

Look how it look just before I restart httpd.



Cloudlinux support suggest maybe its a bug in apache2 2.4.55-1. I do not know what to do, can I safely downgrade apache2?

Thanks

 

[cPRex]

You'd have to monitor the system and see exactly what those processes are and how much traffic is happening then. With just the graph, we'd only be guessing at what could be happening on the machine.

 

[benito]

You'd have to monitor the system and see exactly what those processes are and how much traffic is happening then. With just the graph, we'd only be guessing at what could be happening on the machine.

Yes, I'm monitoring everything. I'll update the ticket when there are enough processes stuck in "Gracefully finishing" so support can see it again. My idea is that they didn't check anything because they didn't consider the load to be a concern. That was only because I had recently restarted apache.

 

[benito]

Hello!

I think my problem is related to this apache mod issue.

Idle H2 threads don't terminate on "graceful" Apache restart (mpm_event) · Issue #212 · icing/mod_h2

Vanilla Apache with "mpm_event" functions in the following way after a "graceful" restart: Old child processes which serve no requests are terminated immediately. Old child processes which have thr...

github.com

I uninstalled mod_https2 in Easyapache to see if the problem persist.

Checking my logs I can see the problem started on January when Apache get updated to Apache/2.4.55

 

[benito]

Hello!

I can confirm that disabling mod_https2 "fixed" the problem. The last 24 hours we have the best load in weeks and no G processes stuck without restarting apache.

I do not know how this works or why this is not happening in my other servers with the mod enabled, but I hope it helps someone else with the same problem.

 

[wavesource]

Confirming this issue also affected my server - running Centos 7. Recurring issue of 5 or 6 nobody processes taking up over 100% CPU each. Kill them, restart Apache, they just reappeared. Examination of what the processes were doing just showed usual low level website activity, serving pages, files etc.

Removed mod_https2 from Apache via EA 4 and problem solved.

Thanks to benito for posting this.

 

[benito]

Confirming this issue also affected my server - running Centos 7. Recurring issue of 5 or 6 nobody processes taking up over 100% CPU each. Kill them, restart Apache, they just reappeared. Examination of what the processes were doing just showed usual low level website activity, serving pages, files etc.

Removed mod_https2 from Apache via EA 4 and problem solved.

Thanks to benito for posting this.

I'm glad I helped. In my case I have even removed http2 on servers that did not have the specific problem of hanging processes. The truth is that on shared servers http2 consumes a lot of resources and I have not had any complaints from any client for not having http2.