Load average counts processes that are running or waiting to run, plus processes in uninterruptible sleep (usually waiting on disk or NFS). Compare it with your CPU count from nproc: a 1-minute load above the core count means work is queueing. Check whether CPU is busy (top shows high us/sy) or idle with high wa — the second means the bottleneck is storage, not CPU.
Symptoms
- uptime or top shows load averages above the number of CPU cores
- SSH sessions and commands feel slow to respond
- Web requests time out while CPU usage looks modest
How to diagnose it
Compare load to core count
The three numbers are 1, 5 and 15-minute averages. A rising 1-minute figure means the problem is current; a high 15-minute figure means it has lasted a while.
uptime nprocSee where CPU time goes
In the %Cpu(s) line, high us or sy means real CPU work; high wa means processes are waiting on I/O; high st means the hypervisor is taking CPU away from your VM.
top -b -n 1 | head -15Watch run queue vs blocked processes
Column r is runnable processes, b is processes blocked on I/O. A large b with low CPU use points at storage.
vmstat 1 5List processes in uninterruptible sleep
Processes in state D count towards load even though they use no CPU. The WCHAN column hints at what they wait for.
ps -eo state,pid,user,wchan:32,cmd | awk '$1 ~ /D/'Find the heaviest CPU users
Sort by CPU to see which processes contribute.
ps -eo pid,user,%cpu,%mem,cmd --sort=-%cpu | head -15
Common causes and fixes
CPU-bound application or runaway process
Identify the process with top or ps, then fix the workload (a loop, a stuck worker, a heavy cron job). Lower its priority while investigating.
sudo renice +10 -p PIDSlow or saturated disk (high iowait)
Use iostat -x to find the device at high %util and the process doing the I/O with iotop. See the high iowait guide.
Hung NFS or network mount
Processes stuck in D state on an unreachable mount drive load up without CPU use. Check mounts and restore connectivity; a lazy unmount can release the hang.
mount | grep -E "nfs|cifs"
sudo umount -l /mnt/shareMemory pressure causing swapping
When RAM runs out the kernel swaps, and swap I/O raises load. Check free -m and the swap guide.
Too many workers for the machine
Reduce PHP-FPM, Gunicorn or queue worker counts so concurrency fits the CPU and memory you have.
Prevent it happening again
- Alert when CPU_loadavg stays above your core count for 15 minutes.
- Alert on CPU_iowait above 20% for 10 minutes to catch storage bottlenecks early.
- Keep load, CPU breakdown and top-process history so you can see what was running when load spiked.
Set it up: Linux monitoring · Ubuntu monitoring.
FAQ
What is a normal load average on Linux?
Anything up to the number of CPU cores is generally fine. A 4-core server at load 3 is busy but not queueing; at load 12 work is waiting.
Why is load high but CPU usage low?
Linux counts processes waiting on disk or network filesystems (state D) in the load average. High load with idle CPU usually means slow storage or a hung mount.
Is load average per core?
No. It is a total for the whole machine, so divide by nproc to compare servers of different sizes.
Related guides
Get alerted before this becomes an outage.
Install the agent in one line. Statance watches CPU, memory, disk and services, and its AI assistant helps find the cause.