"possible run away process"

[likerise]

NOTICE [157.55.35.*:12397-0#APVH_*****.com] No request delivery notification has been received from LSAPI process group [-1], possible run away process.
NOTICE [157.55.35.*:12397-0#APVH_*****.com] Retry with new process group.

I have noticed website downtime and found those log entries all over the place. When this issue happens, every PHP page would take longer and longer time to process and in the end all pages will timed out.

This is a cPanel server with CloudLinux for shared hosting, lots of websites but load is fine, around 1-2 during busy hours, and it shouldn't be a script issue because all php pages, no matter what script users are running, have the same issue.

Static files are all fine while all PHP pages are down.

Switched to Apache for 5 days, working without any issue. If switched to Litespeed, this issue will show up after litespeed running few hours or one day.

Upgraded to lastest verison, still the same.

Can anyone help me which aspect should I look into?

Thank you.

 

[webizen]

what about simple php script like helloworld or phpinfo? this issue recently came up but it was ok before? See if the below helps

http://www.litespeedtech.com/support/forum/showthread.php?t=6405

 

[likerise]

what about simple php script like helloworld or phpinfo?
I did not have a chance to test this, it's a busy server and uptime is really important. This issue did not happen right after I switch to LS so if I want test this I have to wait for another downtime(which I try my best to avoid.)

this issue recently came up but it was ok before?
Just came up. I have been using LS for a long time, and before the issue arise, there is no upgrade made to LS or CL kernel/lve.

See if the below helps
My EP limit is 40 and nPROC limit is 80, while my PHP suEXEC max conn is 20, smaller than both.

I waited a few days until I made this reply, server is running Apache these days, there is no issue during this time(of course load is higher).

 

[webizen]

...
See if the below helps
My EP limit is 40 and nPROC limit is 80, while my PHP suEXEC max conn is 20, smaller than both.
...

PHP suEXEC max conn needs to be bigger than EP limit. Pls raise that or lower the EP limit.

 

[likerise]

Ahh, I thought it should be lower because the #9 reply in the link you gave me:
http://www.litespeedtech.com/support/forum/showthread.php?t=6405

I must set "PHP suEXEC max conn" same as "EP" in cloudlinux ?

>>No, it should be lower than EP.

So that post could be a little misleading.

I will increase PHP suEXEC max conn and test again, thank you for your time!

 

[webizen]

Thanks for pointing out. PHP suEXEC max conn should be LESS than EP which you already set that way. maybe you should lower max conn further.