Our EBS is slow, right? Can I set an EBS to 500M for faster speed and then designate it as the leader? What is the command to designate it?

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

Original topic: 我们那个ebs不是慢么 我可以设定一个ebs为500m 速度快 然后指定他为leader么?指定命令是啥

| username: tidb狂热爱好者

[TiDB Usage Environment] Production Environment / Testing / PoC
[TiDB Version]
[Reproduction Path] What operations were performed when the issue occurred
[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]
Our EBS is slow, right? Can I set an EBS to 500m for faster speed and then designate it as the leader? What is the command to designate it?
There are only 3 TiKVs, one is fast to help with quick writes. The others are slower. Can you provide the specific command?
Does TiDB return based on the first response or does it have to wait for all writes to complete before returning?
Will such changes help improve disk performance and avoid more slow SQL queries?

| username: h5n1 | Original post link

  1. Add evict-leader scheduler in pd-ctl or set the store’s leader weight to 0.
  2. Configure the leader on the specified node using placement-rules in SQL.
| username: tidb狂热爱好者 | Original post link

Thank you for your help.

| username: 小龙虾爱大龙虾 | Original post link

TiDB uses the Raft consensus algorithm to ensure data consistency. For writes, it requires a majority to succeed before returning. In the case of 3 replicas, even if you concentrate the Leader on one node, it still needs to write at least 2 copies to succeed. Personally, I feel this doesn’t significantly improve write performance.

| username: ziptoam | Original post link

I don’t think it will work. It requires strong consistency, and the transaction only ends after writing to three replicas.

| username: zhanggame1 | Original post link

It doesn’t make much sense. For 3 replicas, at least 2 nodes must complete the write for it to be considered complete, so it will still be slowed down by the slower nodes.