Note:
This topic has been translated from a Chinese forum by GPT and might contain errors.Original topic: 什么情况下 ，PD 本身会占用大量 CPU 或 IO 资源？
|
username: 逍遥_猫
Under what circumstances will PD itself consume a large amount of CPU or IO resources?
|
username: 逍遥_猫
|
username: A large number of TSO requests and the generation of scheduling operators
|
username: PD mainly handles these tasks: TSO, processing store and region heartbeats, handling region splits, merges, and migrations. When region information is refreshed, it writes to etcd. TSO writes to etcd once and can pre-allocate for several minutes, so if IO is high, it’s likely due to frequent region changes. TSO doesn’t add much pressure.
|
username: I saw someone in the community reply that this would lead to high CPU usage.
When the PD node has a dashboard and the log search function is used, it will cause high CPU usage.
The dashboard and PD leader are not together by default. In the tiup cluster display, the PD node displaying UI is the dashboard, and L is the leader.
Oh, I saw someone in the community mention that it might be combined. If combined, this will affect CPU usage.
You can set the PD member weight to switch in a specified order.
|
username: The image is not available for translation. Please provide the text you need translated.
|
username: PD should not occupy a large amount of IO resources, but it is possible to occupy a large amount of memory.
|
username: What command should I use to adjust the weight? I see that in our cluster, both the PD leader and the dashboard are together.
PD does not read data, so it should not occupy too much memory.
|
username: This topic was automatically closed 60 days after the last reply. New replies are no longer allowed.