-
Notifications
You must be signed in to change notification settings - Fork 1.1k
Description
Please, answer some short questions which should help us to understand your problem / question better?
- Which image of the operator are you using? e.g. registry.opensource.zalan.do/acid/postgres-operator:v1.9.0
- **Where do you run it - cloud or metal? GKE
- Are you running Postgres Operator in production? [yes]
- Type of issue? [feature request]
Context
I've been running zalando postgresql clusters as the database for Graph node for several months. I collected postgres metrics(from postgres-exporter deployed as a sidecar of postgresql cluster) and patroni metrics also from spilo 8008 port to found the RC of the frequent performance issues (facing when I ran compute-intensive queries in subgraph).
After some investigation, I figured out the reason and it was cpu throttling in postgresql clusters. To make things worse, becaues of slowness by cpu throttling, db connections are increased by graph-node. The increase in the active connection results in the resource usage increase. Cyclic causality
Cpu throttling occurs even if postgres pods don't use much cpu in fact and it turns out k8s cpu limit is ridiculous. I want to stop using cpu limit but when i set it as none, default value 1 is used.
Expectation
If cpu limit field is set as null explicitly, I expect to not set cpu limit to postgresql pods instead of setting default values.
If cpu limit field is not set explicitly, I expect the default limit value.