The fio test results vary greatly

Note:
This topic has been translated from a Chinese forum by GPT and might contain errors.

Original topic: fio测试结果差异很大

| username: jingyesi3401

[TiDB Usage Environment] Production, Testing Environment

[TiDB Version] v5.1.0

[Encountered Problem: Phenomenon and Impact] Preparing to expand the production environment. Before expansion, fio tests were conducted on the new server for verification. However, it is very strange that the IOPS is ridiculously low when using a physical machine with directly mounted raw disks (fdisk + ext4) (the test commands are exactly the same without any changes). The relevant test verification values are as follows:

[Resource Configuration] The server configuration is as follows:

Note: The test environment and test records are as follows:
测试环境.txt (3.6 KB)
测试记录.txt (9.8 KB)
机器信息.txt (461 bytes)

| username: Billmay表妹 | Original post link

IOPS not being high might be due to low pressure on the disk, which could be caused by bottlenecks in the application layer, TiDB server, or TiKV.

You can first check the CPU utilization of several threads in TiKV. If it’s a TiKV issue, you can refer to the documentation https://docs.pingcap.com/zh/tidb/v4.0/tune-tikv-thread-performance for tuning.

| username: Hacker_nTkchcIf | Original post link

You can first try a raw disk stress test without partitioning.

| username: jingyesi3401 | Original post link

The bare disk test results are as follows: IOPS=7622

| username: yilong | Original post link

  1. What was the data from the previous test?
  2. Is it the exact same operating system version and using the same mount parameters?
  3. If there are previous test standards, you can test according to the previous configuration.
| username: system | Original post link

This topic was automatically closed 60 days after the last reply. New replies are no longer allowed.