Issues with Resource Control Prioritization

Translated
1

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

Original topic: resource_control优先级的问题

| username: TiDBer_o0MXVuK4

[TiDB Usage Environment] Poc
[TiDB Version] 6.1.2
[Problem Encountered] The resource_control configured in global overrides the resource_control at each service level in server_configs. Shouldn’t it be the other way around?
[Resource Configuration]

global:
  user: hadoop
  group: hadoop
  ssh_port: 22
  ssh_type: builtin
  deploy_dir: /home/hadoop/apps/em-deploy
  data_dir: /home/hadoop/apps/tidb-data
  resource_control:
    memory_limit: 112G
  os: linux
  arch: amd64
monitored:
  node_exporter_port: 9100
  blackbox_exporter_port: 9115
  deploy_dir: /home/hadoop/apps/em-deploy/monitor-9100
  data_dir: /home/hadoop/apps/tidb-data/monitor-9100
  log_dir: /home/hadoop/apps/em-deploy/monitor-9100/log
server_configs:
  tidb:
    log.level: info
    oom-use-tmp-storage: true
    performance:
      memory-usage-alarm-ratio: 0.8
      txn-total-size-limit: 10737418240
    resource_control:
      memory_limit: 112G
    tmp-storage-path: /home/hadoop/apps/tidb-data/tidb-temp
    tmp-storage-quota: 26843545600
  pd:
    log.level: info
    replication.enable-placement-rules: true
    resource_control:
      memory_limit: 12G
    schedule.leader-schedule-limit: 4
    schedule.region-schedule-limit: 2048
    schedule.replica-schedule-limit: 64

[PD’s systemd configuration]

2
21139×523 50.9 KB

[TiDB documentation screenshot]
1
1848×110 21.1 KB

2
| username: Jiawei | Original post link

It really seems like it. You can raise an issue and see.

3
| username: TiDBer_o0MXVuK4 | Original post link

The issue has been resolved. Currently, it does not support direct configuration of resource control at the component level. It can either be global or specific to a particular service node, such as a specific TiKV node. This makes configuration inconvenient in larger clusters. Hopefully, in the future, it will be possible to configure by component or by group (for example, if some components have two batches of machines with different configurations, labels can be assigned to the nodes, dividing them into two groups for separate control).

4
| username: Jiawei | Original post link

I have already reported this to the official team. It is indeed a minor bug, and I have confirmed with them that it will be optimized in the future.

5
| username: TiDBer_o0MXVuK4 | Original post link

Alright, it would be great if it can be fixed.

6
| username: system | Original post link

This topic was automatically closed 60 days after the last reply. New replies are no longer allowed.