"Why Is Load 10 but CPU 0%?"
Think of a hospital emergency room. The "load" on the ER isn't just the number of doctors actively treating patients — it's the total number of people who need attention right now, including those sitting in the waiting room. A doctor can be completely idle while the waiting room is packed, because the bottleneck isn't the doctor — it's the MRI machine that everyone is waiting for.
Linux Load Average works the same way. When you run uptime or top, you see three numbers, e.g. load average: 1.23, 0.87, 0.54. These represent the average number of processes in one of these states over the last 1 minute, 5 minutes, and 15 minutes:
- Running (R): Actively executing on a CPU core.
- Runnable (R): Ready to run, waiting for a free CPU core.
- Uninterruptible Sleep (D): Waiting for hardware I/O to complete — cannot be interrupted by signals.
The D-state is the critical insight. A process waiting for a disk read that takes 30 seconds contributes 1.0 to the load average for those 30 seconds, even though it uses 0% CPU. If 50 processes are all waiting on a broken NFS mount, your load average is 50 — and your CPU is completely idle.
Reading load average in context:
Load average is only meaningful relative to the number of CPU cores — checkable with nproc. A load of 4.0 on a 4-core machine means the system is fully saturated. The same load of 4.0 on a 32-core machine means it's running at 12.5% capacity.
Rule of thumb:
- Load / Core Count < 1.0 → System has headroom
- Load / Core Count ≈ 1.0 → System is at capacity
- Load / Core Count > 1.0 → Processes are queuing; investigate immediately
The 1-minute, 5-minute, and 15-minute trend tells you whether the situation is improving or worsening. A load of 8.0, 4.0, 2.0 (rising) is more alarming than 8.0, 12.0, 15.0 (falling).