Issues with TiKV Scaling

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

Original topic: tikv扩容问题

| username: TiDBer_yUoxD0vR

[TiDB Usage Environment] Production Environment / Testing / PoC
[TiDB Version] v3.0.13
[Reproduction Path] What operations were performed when the issue occurred
When expanding TiKV nodes in TiDB v3.0.13, can you expand another node before the first node has balanced?
With the parameters region-schedule-limit and leader-schedule-limit unchanged, does expanding one TiKV node at a time have less impact on the business, or does expanding multiple nodes simultaneously have less impact?
[Encountered Issue: Issue Phenomenon and Impact]
[Resource Configuration] Go to TiDB Dashboard - Cluster Info - Hosts and take a screenshot of this page
[Attachments: Screenshots/Logs/Monitoring]

| username: xfworld | Original post link

It doesn’t matter if you expand it casually, as long as the scheduler is not the region leader, it won’t have much impact.

| username: Kongdom | Original post link

I think expanding multiple times at once is faster, and the overall cycle is shorter.

| username: tidb菜鸟一只 | Original post link

Adding one TiKV node at a time has a smaller impact on the business because with parameters like region-schedule-limit and leader-schedule-limit unchanged, expanding from 3 nodes to 4 nodes means the data (including leaders) on the 3 nodes changes from 1/3 to 1/4. Expanding from 3 nodes to 5 nodes means the data (including leaders) on the 3 nodes changes from 1/3 to 1/5. Theoretically, the scheduling time is longer, and more leaders need to be moved. However, once the scheduling is complete, the performance of 5 nodes is definitely better than that of 4 nodes.

| username: 像风一样的男子 | Original post link

It is recommended to increase the parameters to speed up region migration during business downturns.

| username: zhanggame1 | Original post link

It’s best to add them all at once, otherwise adding them one by one requires balancing twice.