When CPU Is Low, but Load Average Is High

Written By Krishna Chaitanya Kathari

Load average represents the number of processes that are:

  • Running on the CPU, or

  • Waiting to run because they are blocked (most commonly on I/O)

So:

Load average ≠ CPU usage

A system can have idle CPUs and still be overloaded.

The Most Common Cause: Disk I/O Wait

When processes are blocked on disk reads or writes:

  • CPU stays idle

  • Load average increases

  • The system feels slow

This happens because processes are stuck in uninterruptible sleep (D state) and cannot be scheduled.

How to Confirm It Quickly

Run:

vmstat 1

Example output (simplified):

procs -----------memory---------- -----cpu-----
 r  b   free   cache   us  sy  id  wa
 1  6  82000  450000    4   3  68  25

What to focus on:

  • us (user CPU) → application work

  • sy (system CPU) → kernel work

  • wa (I/O wait) → CPU idle, waiting on disk

Here:

  • us + sy is low

  • wa is high
    → CPU is available, but processes are blocked.

Identify the Disk Bottleneck

Next, run:

iostat -x 1

Example output:

Device            r/s   w/s   rkB/s   wkB/s  await  avgqu-sz  %util
nvme0n1          120    85   4800     6200    45.3     7.8     98.6

Red flags:

  • High await → slow I/O response

  • High avgqu-sz → requests piling up

  • High %util (~100%) → disk fully saturated

This confirms disk I/O is the bottleneck, not CPU.

Key Takeaway

Low CPU + High Load = Something else is blocking progress

Most of the time, that “something” is disk I/O.

Understanding this distinction prevents chasing the wrong problem
and saves real incident response time.

One Rule to Remember

CPU usage shows work being done.
Load average shows work waiting to be done.

Krishna Chaitanya Kathari