Note:
This topic has been translated from a Chinese forum by GPT and might contain errors.
Original topic: 新人tidb部署问题，大家帮帮忙看看
Deploying TiDB on a virtual platform, an error occurred indicating that the disk check failed. It is an SSD disk, formatted as EXT4 with the required parameters according to the deployment requirements. Is there anything else that needs to be set up?
fatal: [10.137.100.114]: FAILED! => {“changed”: false, “msg”: “fio: randread iops of tikv_data_dir disk is too low: 646 < 40000, it is strongly recommended to use SSD disks for TiKV and PD, or there might be performance issues.”}
fatal: [10.137.100.115]: FAILED! => {“changed”: false, “msg”: “fio: randread iops of tikv_data_dir disk is too low: 572 < 40000, it is strongly recommended to use SSD disks for TiKV and PD, or there might be performance issues.”}
fatal: [10.137.100.116]: FAILED! => {“changed”: false, “msg”: “fio: randread iops of tikv_data_dir disk is too low: 593 < 40000, it is strongly recommended to use SSD disks for TiKV and PD, or there might be performance issues.”}
The fio disk performance test failed. You can skip the check when installing the test cluster, but for production, it’s better to use a better machine.
It’s in production, so it’s not easy to skip. It’s deployed on a virtual cloud platform, so there shouldn’t be any performance issues.
The difference in IOPS is quite significant.
Yeah, I’m also puzzled. Why is there such a big difference? Is there something wrong? 
Let the company’s infrastructure department check if there is an IOPS limit or test if the hardware performance is just that poor.
I asked the virtual platform person to check, and they said there is no IOPS limit.
I used fio for testing, and the IOPS are sufficient.
Disk stats (read/write):
sdb: ios=339118/144466, merge=0/39, ticks=1142752/51090, in_queue=1193819, util=99.97%
You can force the deployment. You can also deploy on a separate disk by purchasing a local hard drive for each machine.
By default, disk performance is not checked. Did you add the --enable-disk option?
Please provide the command for installing TiUP.
An error occurred while initializing the system environment with ansible-playbook bootstrap.yml.
I referred to 使用 TiDB Ansible 部署 TiDB 集群 | PingCAP 归档文档站
document to deploy V3.0
and encountered an error when initializing the system environment.
Use a newer version and deploy directly with tiup.
Which version is good to use now, and is it compatible with the lower version? Because I still need to use a lower version, that’s why I want to install an old version. There are only three hosts.
Just use tiup to install the new version. It seems that version 5.4 is more commonly used in production.
How about changing the virtual device node from “SCSI Controller” to “NVMe Controller” and then testing fio again?
This topic was automatically closed 60 days after the last reply. New replies are no longer allowed.
|
username: TiDBer_gypkMA1c
|
username: 
|
username: 
|
username: 
|
username: 
|
username: 
|
username: 
|
username: 
|
username: