Note:
This topic has been translated from a Chinese forum by GPT and might contain errors.
Original topic: 云上tidb数据扩容各位是愿意扩容ebs后重启，还是新建ecs后加tikv进去？那种方式对tidb影响小
[TiDB Usage Environment] Production Environment / Testing / PoC
[TiDB Version]
[Reproduction Path] What operations were performed that led to the issue
[Encountered Issue: Issue Phenomenon and Impact]
[Resource Configuration] Enter TiDB Dashboard - Cluster Info - Hosts and take a screenshot of this page
[Attachments: Screenshots/Logs/Monitoring]
When expanding TiDB data on the cloud, do you prefer to expand EBS and then restart, or create a new ECS and add TiKV? Which method has less impact on TiDB?
Do you recommend using local disks or EBS cloud disks?
One is vertical scaling, and the other is horizontal scaling. Horizontal scaling has less impact on usage and does not interrupt the business. Vertical scaling is better for hotspot issues; if hotspots are still difficult to resolve, upgrading the node specifications can handle all requests more effectively.
This also depends on the disk size. I think the optimal size for a KV is 2TB; if it exceeds that, add a node. Expanding the disk is the fastest and least noticeable method. Adding a node can affect database performance if the region scheduling speed is too fast.
Horizontal scaling can better distribute hotspots.
The immovable hotspot can only be handled by a large-scale single node.
Personally, I still recommend operating TiKV nodes. Firstly, it does not affect the business. Secondly, it increases computing resources. Thirdly, it will be convenient for scaling down in the future.
I feel that the method of “adding TiKV after creating a new ECS” is better.
The main reason my previous company used the method of “expanding EBS and then restarting” was because of budget constraints; adding storage was cheaper than expanding a machine. 
Adding only storage won’t alleviate system pressure; adding TiKV nodes is the key.
I personally prefer to create new ECS for horizontal scaling.
The impact of the action itself is basically the same.
Personally, I feel that horizontal scaling is more stable, which is the advantage of distributed systems. When storage is insufficient, adding nodes allows for automatic data scheduling.
Add TiKV after creating a new ECS.
|
username: tidb狂热爱好者
|
username: 
|
username: 
|
username: 
|
username: 
|
username: 
|
username: 
|
username: 
|
username: 
|
username: 
|
username: