Will deploying PD and Prometheus on the TiDB-server affect the overall performance of the cluster?

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

Original topic: pd和prometheus在tidb-server服务器上部署,会不会影响到集群的整体性能?

| username: xiaoxiaozuofang

[TiDB Usage Environment] Production Environment
[TiDB Version] tidb v6.1.0
[Reproduction Path] What operations were performed when the issue occurred
[Encountered Issue: Issue Phenomenon and Impact] Will deploying PD and Prometheus on the tidb-server affect the overall performance of the cluster?
[Resource Configuration]
[Attachments: Screenshots/Logs/Monitoring]

| username: 我是咖啡哥 | Original post link

It is recommended to deploy separately, a virtual machine will also work.

| username: zhanggame1 | Original post link

It’s best to separate them. PD and monitoring don’t consume much resources, so you can deploy them on machines with average performance. TiDB nodes, on the other hand, are very CPU and memory intensive.

| username: DBRE | Original post link

Prometheus may cause a CPU spike due to a large PromQL query, so it is recommended to deploy it separately.

| username: 秋季番Oo | Original post link

When the data volume of Prometheus is relatively large, or executing some computationally complex PromQL queries, it will consume a lot of CPU. It is better to deploy them separately in a production environment. After all, PD is very important for the entire TiDB cluster.

| username: tidb菜鸟一只 | Original post link

Yes, if there are available resources, it is definitely more reasonable to deploy them separately.

| username: 啦啦啦啦啦 | Original post link

If resources are insufficient, you might consider mixed deployment on non-leader PD nodes, as the impact would be relatively small.

| username: buptzhoutian | Original post link

There is not much problem with deploying PD and TiDB together; however, in a production environment, it is best not to deploy Prometheus on other nodes of the cluster. For example, when Prometheus local storage is in Compaction, it may occupy too many disk resources, and PD is very sensitive to disk write latency.

| username: system | Original post link

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