Expected High Load Average on Cloudlinux OS with kmod-lve v. 2.1-56+

Issue

After updating kernel and kmod-lve on a CloudLinux server, the system started reporting very high load average while CPU, IO, vmstat show do not indicate the issue with load. 
System remains responsive. However, /proc/loadavg shows high value while there is quite a low amount of running tasks.

Environment

• Cloudlinux OS
• kmod-lve v. 2.1-56+

Solution

Our kernel developers examined the situation and figured out that the higher load average is expected and these are the real numbers.

kmod-lve versions prior to 2.1-56 followed a different method of calculating the load average - the load created by tasks from non-root cgroups (such as jetbackup) was not counted. 
This bug has been addressed in the newer kmod-lve versions.

The current LA accounts for:

1. any amount of tasks (running/uninterruptible) in LVE as 1.0;
2. any other tasks as N.0 where N is an amount of tasks that are running or uninterruptible.

If you still experience an issue with the aforementioned calculation method, or if the LA is higher than it should be considering the expected behavior, or if your web services are affected by it, please do not hesitate to open a ticket with Cloudlinux Support Team for further investigation.

Cause

kmod-lve versions prior to v. 2.1-56 underestimated load average, and the newer versions contain a bug fix for it.