TiDB Disk Quota

Translated
1

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

Original topic: TIDB磁盘配额

| username: 渲染888

Currently planning to deploy a TiDB cluster using 9 ECS instances (3 for both TiDB and PD, 2 for TiKV, 3 for TiFlash, and 1 for monitoring). How should the disk allocation for each ECS be configured? The data volume is less than 1TB.

2
| username: hey-hoho | Original post link

It’s a bit strange to have 2 TiKV nodes. If you plan to use 3 replicas, you should start with at least 3 TiKV nodes, right?

3
| username: 渲染888 | Original post link

It’s written incorrectly, it should be 3 TiKV and 2 TiFlash.

4
| username: ddhe9527 | Original post link

If the data volume is calculated as 1TB, three replicas would be 3TB. Considering some space amplification and old data not yet cleaned by GC, each of the three TiKV instances should plan for at least 1.5TB. The disk capacity for TiFlash depends on the size of the tables for which you want to create TiFlash replicas. Disk planning for other components can refer to the following documentation:

5
| username: cs58_dba | Original post link
  • It is not recommended for the storage of a TiKV node to exceed 2TB, otherwise, there will be too many regions in a single store.
    • For a 3TB instance with 3 replicas, there will be a total of 9TB of data, distributed across 5 TiKV nodes, so the storage data of a single TiKV will be less than 2TB.
    • Generally, when planning, the disk is fixed at around 2TB, and when the single instance disk is almost full, just expand the nodes.
  • The recommended TiKV hard disk size configuration is PCI-E SSD not exceeding 2TB, and ordinary SSD not exceeding 1.5TB.
  • TiFlash supports multi-disk deployment.
6
| username: system | Original post link

This topic will be automatically closed 60 days after the last reply. No new replies are allowed.