CloudLinux Bug: Please, read if you are experiencing problems

[adam.heavens]

but with order of lvectl execution

[iseletsk]

There is a bug in the order /etc/init.d/lvectl is executed. It is executed after apache (httpd) is. As the result if any site gets accessed before lvectl is started, the LVE for that site is set with default limits. The CPU limit is set by default to 100% / 10 * number of cores.
This means that if you have 8 cores -- cpu limit will be set to 1, and site will not work. One of the side effects of it is that load shuts up very high right after the start, keeps going up, even on the server that is not loaded at all.
To see if you are experiencing the problem, run
cat /proc/lve/list
If you see LVEs with CPU limit not being set at default value (or value you have set).

To resolve the issue:
Edit /etc/init.d/lvectl
Change line # chkconfig: 2345 99 90
to line: # chkconfig: 2345 11 90
execute:
chkconfig lvectl resetpriorities

reboot the server.

We should have updated package that should resolve this issue fairly soon. I will update the post once we do.

[guillermocalvo]

Igor:
There is a typo

chkconfig vectl resetpriorities should be chkconfig lvectl resetpriorities

[iseletsk]

> Guillermo Calvo wrote:
> Igor:
>
> There is a typo
>
> chkconfig vectl resetpriorities should be chkconfig lvectl resetpriorities

Thanks a lot.

[iseletsk]

I just confirmed that we are not counting things correctly with mod_fcgid. I will try to have a fix by Monday.

[iseletsk]

The fix for the order in which lvectl is executed is in repository right now.
yum update lve-utils should fix the issue (you might need to reboot).
You should get version 0.3-0.1 installed.