prometheus-operator VS k8s-prometheus-adapter

The project repository for Prometheus-operator can be found here, The repo defines the CRDs and the controller. You can follow this documentation for the installation. which will require the creation of metrics exporters, node exporters, scrape configurations, etc.

Personally I use kubernetes, k3s is kind of lightweight, with the Prometheus operator.

https://k3s.io/

https://prometheus-operator.dev/

Kubernetes is not for everyone and is far from perfect but you already use Docker and you seem to seek many features offered by Kubernetes.

I see that the operator is a community-run project. Does anybody have any experience (positive or negative) on running the operator itself? I wonder if it is more for multi-tenant Prometheus. For single tenant Prometheus collection, I guess the big gain is abstracting scrape config into ServiceMonitor and PodMonitor resources? Is there anything else that this makes much better?

But in all honesty if you’re on k8s you’re probably best served by the Prometheus operator. It’ll give you all that and more, easily.

Try some monitoring.

It’s a common pattern [1] how else can users override the child objects you’re controller creates?

1 https://github.com/prometheus-operator/prometheus-operator/b...

The Prometheus operator is this one https://github.com/prometheus-operator/prometheus-operator

At a minimum, you want to look at your metrics using the Prometheus Operator. You can write alerts for resource requests.

Prometheus Operator is an independent project from the Prometheus project. I know, it can lead to confusion. In the official README you can find short comparison. Basically, Prometheus Operator does what an operator should do - provides Kubernetes native deployment and management of Prometheus and related monitoring components like Grafana or Alert Manager.

The last one is mostly an observability stack with Prometheus, Metric server, and Prometheus adapter to have excellent insights into what is happening on the cluster. You can reuse the same stack for autoscaling by repurposing all the data collected for monitoring.

I'm deploying kube-prometheus-stack along with prometheus-adapter in my monitoring stack for custom metrics.

I would like to see if anyone deployed prometheus-adapter and kube-prometheus-stack together for monitoring?

For us it is saturation of CPU and thread pool. It's implemented by exposing metrics of the thread pool to prometheus and turning that into a custom metric. (see) Looking at scaling based on job queue length next.

If you are using Prometheus or kube-prometheus-stack, you will need https://github.com/kubernetes-sigs/prometheus-adapter We are using it to scale our Pods based on number of messages in RabbitMQ queue. There also a walkthrough on https://github.com/kubernetes-sigs/prometheus-adapter/blob/master/docs/walkthrough.md

A prometheus-adapter installation.

Prometheus adapter to get custom/external metrics from Prometheus instance into Kubernetes API.

I never tested this but you have customized metrics API if the value % is available should work from my point of view Check this here https://github.com/kubernetes-sigs/prometheus-adapter

CPU and memory might not be the right metrics for your application to make scaling decisions. In such cases, you can use HPA (or VPA) with custom metrics as an alternative. To use custom metrics for autoscaling, you can use a custom metrics adapter instead of the Kubernetes Metrics Server. Popular custom metrics adapters are the Prometheus adapter and Kubernetes Event-Driven Autoscaler (KEDA).