Note:
This topic has been translated from a Chinese forum by GPT and might contain errors.Original topic: 新加一个tiflash节点，原来旧的tiflash region不能自动均衡到新的tiflash上么？
|
username: lindoubled
Adding a new TiFlash node, I found that the old TiFlash regions do not automatically migrate to this new TiFlash node.
Version: 5.3.0
TiFlash probably doesn’t have the same automatic balancing mechanism as TiKV.
Set the TiFlash replica to the corresponding number of TiFlash instances for high availability. If you need to migrate replicas to other TiFlash instances, you need to do it manually. enable-for-tiflash is a switch that controls whether hotspot scheduling is effective for TiFlash. It is usually turned on, and if turned off, there will be no hotspot scheduling between TiFlash instances.
Yes, there is. I adjusted the weight of TiFlash, and the region was immediately scheduled, migrating from one TiFlash to another.
I enabled the shuffle-region-scheduler scheduling, and TiFlash can automatically balance.
enable-for-tiflash can be set using pd-ctl. It is generally true by default.
shuffle-region-schedule schedules the learner.
Okay, it seems that there isn’t much documentation about shuffle-region-schedule. What is it mainly used for?
scheduler add shuffle-region-scheduler: Randomly schedule Regions on different stores
Supports learner
Will it keep scheduling, or will it stop scheduling once balanced?
After completion, it should not be continuously scheduled. You can use
scheduler show to check.
balance-region-scheduler: Keeps the Peers balanced across different nodes.
Yes, including all peers. I remember that the previous shuffle-region-schedule did not support learners. Which version was upgraded?
The region-balance-scheduler will score based on…
The region-balance-scheduler is created by default, so in the scenario described by the original poster, I understand that TiFlash Regions should be automatically scheduled to the new node without needing to create the shuffle-region-scheduler. Is that correct?
Yes, it looks like he has mixed two TiFlash instances on the same machine.
This doesn’t seem to be related, hybrid deployment is quite normal.
|
username: 
|
username: 
|
username: