Best way for monitoring AKS

5/19/2020

I need to find and compare the best solution for monitoring AKS.

Exist any comparison between different products? I dont find any link

I need to compare about pricing, functionalities, etc

We tried with Log analytics but is very expensive.

-- exitista
azure
azure-aks
azure-kubernetes

1 Answer

5/19/2020

While Product recommandation isn't something you will get on StackOver Flow.

I would recommend you to have a look at the OSS stack for properly monitoring AKS and Kubernetes in general. This solution will work on any Kubernetes cluster (AKS/EKS/GKE/BareMatel).

Start with the Prometheus Operator, this will bring in Grafana/Prometheus/AlertManager and a set of default dashboard and alert for your Kubernetes Cluster.

https://github.com/helm/charts/tree/master/stable/prometheus-operator

You even get monitoring for the control plane:

kube-apiserver
kube-scheduler
kube-controller-manager
etcd
kube-dns/coredns
kube-proxy

For better storage (since Prometheus is meant to keep a short amount of retention) have a look at configuring your stack with Thanos: https://thanos.io

This will allow you to augment your retention of metrics to almost an unlimited amount.

As far as Vendors goes, a lot of vendor will have the same price when you start to deal with them. Some of them will rely on their own agents to be installed while other will rely on Prometheus and Kube-State-Metrics to be installed.

While metrics are great, you should allow you user to have access to traces, this help identifying the flows and bottle neck of the different sessions.

https://www.jaegertracing.io/

https://www.jaegertracing.io/docs/1.18/operator/

https://github.com/jaegertracing/helm-charts

Finally for Log and Log Indexing, the ELK stack is your Go To Solution.

https://github.com/elastic/cloud-on-k8s

The Elastic team have been working on a good operator to facilitate the management of the ELK cluster on Kubernetes.

-- djsly
Source: StackOverflow