iostat Command in Linux: Monitor Disk and CPU I/O

Quick take: iostat -x 2 shows extended per-device disk statistics refreshed every 2 seconds. Key metrics are %util (device busy percentage), await (average I/O wait time in ms), rMB/s and wMB/s (read/write throughput).

Introduction

iostat (I/O Statistics) is part of the sysstat package and reports CPU utilization and disk I/O statistics per device. It is the standard tool for diagnosing whether poor application performance is caused by a disk bottleneck — too many requests queuing, high latency, or a device running at capacity.

Without iostat, you might spend hours tuning application code or database queries when the real problem is a spinning disk saturated at 95% utilization. iostat tells you in seconds.

Install iostat

# Ubuntu / Debian
sudo apt install sysstat

# RHEL / CentOS / Rocky
sudo dnf install sysstat

# Verify
iostat -V

Syntax

iostat [OPTIONS] [device] [interval] [count]

interval is the refresh interval in seconds. count is how many reports to print before exiting. Omitting both prints one cumulative report since boot.

Common Options

Practical Examples

# Basic report — all devices since boot
iostat

# Extended stats, refresh every 2 seconds
iostat -x 2

# Extended stats in MB/s, every 5 seconds
iostat -xm 5

# Monitor a specific device
iostat -x sda 2

# Monitor with device mapper names (shows LVM volume names)
iostat -xN 2

# Show only disk stats (no CPU) with timestamps
iostat -dtm 5

# 10 reports at 2-second intervals then stop
iostat -x 2 10

# Show partitions
iostat -p sda -x

Reading iostat Output

The basic output section:

avg-cpu:  %user   %nice %system %iowait  %steal   %idle
           2.34    0.00    0.87    12.45    0.00   84.34

%iowait is the most important CPU metric: percentage of time the CPU was idle while waiting for I/O. Values consistently above 20–30% indicate a disk bottleneck stalling the CPU.

The extended device section (-x):

Device  r/s  w/s  rMB/s  wMB/s  await  r_await  w_await  %util
sda    45.2  23.1   2.18   1.45   8.23     5.12    14.67  68.4
sdb     2.1   0.8   0.10   0.04   1.05     0.90     1.32   0.8

Key columns explained:

• r/s, w/s — reads and writes per second (IOPS)
• rMB/s, wMB/s — throughput in megabytes per second
• await — average total I/O completion time in milliseconds (queue + service)
• r_await, w_await — separate read and write await times
• %util — percentage of time the device was busy; near 100% means saturation

Diagnosing Disk Problems

High %util (>80%) with high await: The disk is saturated. Either the workload exceeds disk capacity or a specific process is monopolizing I/O. Check with iotop to find the process.

# Find which process is doing the most I/O
sudo iotop -o

High %iowait in CPU report: Application threads are stalling waiting for disk. This is a disk bottleneck in the CPU's view — the CPU is available but waiting on storage.

High await but low %util (on SSDs): May indicate I/O queue depth issues or filesystem overhead. Check aqu-sz (average queue size) — consistently high queues on SSDs suggest driver or queue configuration problems.

Final Thoughts

Make iostat -xm 2 your default disk monitoring command. The combination of %util, await, and throughput metrics tells you immediately whether you have a disk bottleneck, and if so, which device and in which direction (read vs write). Pair it with iotop to identify the specific process driving the load.