Prometheus requirements for the machine's CPU and memory

[lwx294821]

How much memory and cpu are set by deploying prometheus in k8s?

[Ghostbaby]

resources:
    limits:
      cpu: 5000m
      memory: 50Gi
    requests:
      cpu: 3000m
      memory: 400Mi

[lwx294821]

@Ghostbaby thanks. The official has instructions on how to set the size?

[metalmatze]

I don't think the Prometheus Operator itself sets any requests or limits itself:
https://github.com/coreos/prometheus-operator/blob/04d7a3991fc53dffd8a81c580cd4758cf7fbacb3/pkg/prometheus/statefulset.go#L718-L723

However, in kube-prometheus (which uses the Prometheus Operator) we set some requests:
https://github.com/coreos/kube-prometheus/blob/8405360a467a34fca34735d92c763ae38bfe5917/manifests/prometheus-prometheus.yaml#L19-L21

[lwx294821]

thanks @metalmatze

[dioguerra]

I did some tests and this is where i arrived with the stable/prometheus-operator standard deployments

RAM:: 256 (base) + Nodes * 40 [MB]
CPU:: 128 (base) + Nodes * 7 [mCPU]
Disk:: 15 GB for 2 weeks (needs refinement)

I'm still looking for the values on the DISK capacity usage per number of numMetrics/pods/timesample
Also, on the CPU and memory i didnt specifically relate to the numMetrics. The high value on CPU actually depends on the required capacity to do Data packing.

This memory works good for packing seen between 2 ~ 4 hours window.

Would like to get some pointers if you have something similar so that we could compare values.
I tried this for a 1:100 nodes cluster so some values are extrapulated (mainly for the high number of nodes where i would expect that resources stabilize in a log way)

[stale]

This issue has been automatically marked as stale because it has not had any activity in last 60d. Thank you for your contributions.

[paulfantom]

This issue hasn't been updated for a longer period of time. Since then we made significant changes to prometheus-operator. If you think this issue is still valid, please reopen it.