What Is the Load Average in Linux and How to Check It

What is the load average

The load average is the average number of processes that wanted the CPU over a period. It counts two kinds of processes:

1. Runnable. They run now or wait in line for a CPU.
2. Uninterruptible. They wait for something such as disk input and output, and cannot be interrupted.

It is a number, not a percentage.

The three values

1. 1 minute. The very recent load.
2. 5 minutes. A short trend.
3. 15 minutes. The longer trend.

Example: 0.50, 1.20, 2.40 means the load is falling. 3.50, 1.20, 0.40 means it is rising.

How to read the values

• Compare the load with the number of CPU cores. Find the cores with nproc.
• A load equal to the core count means the CPU is fully used. On a 4-core server, a load of 4.00 is full.
• A load far above the core count means processes wait in a queue.
• A load of 6 is fine on 8 cores but a problem on 2 cores.
• The type of work matters. Tasks like video encoding raise the load naturally.

Check the load average

uptime

uptime

The last three numbers are the 1, 5 and 15 minute values.

top

top

The load average is in the first line.

htop

htop

/proc/loadavg

cat /proc/loadavg

The first three numbers are the load averages. The fourth is running/total processes, and the last is the most recent PID.

High load but low CPU use

Then processes probably wait for the disk or the network. Check with iostat -x 2 or vmstat 2 and look at the I/O wait.

What to do when the load is high

1. Find the heaviest processes with top or ps aux --sort=-%cpu | head.
2. Check disk and memory use.
3. Stop or tune the process, or add CPU cores.

Quick recap

• Load average has three numbers: 1, 5 and 15 minutes.
• Compare it with the number of CPU cores.