tikv / tikv/pd

Scatter will not work when there are some unhealthy store in the cluster

Open
#5,484 0 comments 0 reactions 0 assignees View on GitHub
type/enhancement
Dominant language
Go
Stars
1.2k
Forks
783
Avg merge
5d 21h
Merged PRs (30d)
36

Description

## Enhancement Task
Split table 1000 after downing the tikv 10min, the disribution of the replica will not balance well.
![origin_img_v2_d1c8f40e-d845-4036-99d4-1ea4c060503g](https://user-images.githubusercontent.com/23159587/189260200-413b4494-b409-436a-89fd-72bd0c565b9d.jpg)

the root cause:
1. scatter always pick the store who has the less region.
2. operaoter builder will failed if some checker failed like the store is unhealthy.
3. the next region will pick the unhealty store again, so all the region will failed.

some monitor may mistake:
image

the original region peer will recount many time if the operator build failed:
https://github.com/tikv/pd/blob/8e92dcde2e52dc04ace61f37d6ea6c83a0a97c96/server/schedule/region_scatterer.go#L363-L371

Contributor guide

Open the contributing guide

Assessment

This issue has not been assessed yet.

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.