使用iostat测量HDD性能结果偏低，疑操作/参数有误求助

Hey there! Let's walk through why your iostat readings for your /home disk are coming in unexpectedly low, and how to fix your measurement setup.

1. First, Confirm You're Targeting the Correct Disk #

You mentioned your system is on sda and /home is on another disk—let's make sure you're monitoring the right device. Run this command to find exactly which disk /home is mounted on:

df -h /home

Look at the Filesystem column in the output. It might be something like /dev/sdb, /dev/sdc, or even an LVM logical volume (e.g., /dev/mapper/vg_home-lv_home). If it's an LVM volume, you'll need to monitor the underlying physical disk(s) (use lvdisplay and pvdisplay to trace back to the physical device) instead of the logical volume itself—iostat won't give accurate raw disk stats for LVM volumes directly.

2. Fix Your iostat Parameters #

The default iostat output only shows CPU stats, and even when you enable disk stats, it gives average values since system boot—which will be low if the disk is mostly idle. Use these parameters to get real-time, meaningful data:

• -d: Focus only on disk stats (skip CPU)
• -x: Show extended stats (including IOPS, read/write throughput)
• Add an interval and count (e.g., 5 3) to get repeated snapshots every X seconds, instead of boot-time averages

Example command:

iostat -dx 5 3

Key metrics to focus on:

• tps: Total IO operations per second (your IOPS)
• rkB/s: Read throughput in kilobytes per second
• wkB/s: Write throughput in kilobytes per second

3. Ensure the Disk is Under Load #

If your disk isn't actively reading/writing, iostat will show low numbers by default. Generate some test load in a separate terminal while running iostat:

• For a read test:

  # Create a test file first if you don't have large files on /home
  dd if=/dev/zero of=/home/test_read_file bs=1M count=1000
  dd if=/home/test_read_file of=/dev/null bs=1M
  

• For a write test:

  dd if=/dev/zero of=/home/test_write_file bs=1M count=1000 conv=fdatasync
  

(Don't forget to delete the test files afterward with rm /home/test_read_file /home/test_write_file)

4. Rule Out Disk Hardware/Health Issues #

If you've confirmed the right device, used correct parameters, and generated load but still see low stats, check the disk's health:

smartctl -a /dev/your_disk_device

Look for warnings in the SMART data—things like high bad sector counts, degraded cache performance, or signs of disk aging could cause poor throughput.

内容的提问来源于stack exchange，提问作者Tomek