Prerequisites

1. Kubectl
   • The cli that will be used to work with any of the K8s clusters
2. Gcloud sdk
   • The cli that will be used to work with Google Cloud in order to provision and delete resources
3. Helm
   • Allows us to automate and better control our kubernetes resources. Helms allows you to define a custom application as a resource and it will manage all the subcomponents for you.

Overview

Deployment and configuration is done via Helm. The reason for the decoupled configuration/charts is to have better control over each component. The full stack chart for Loki + Grafana + Prometheus + Promtail is really restricted in terms of configurations, like importing custom dashboards during provisioning. cAdvisor is used to expose some additional metrics for container resource usage.

Managing the GKE cluster

./cluster.sh <cluster-name> <project-name> - To build,delete or get credentials for your cluster - cluster-name - Name you want to give to the new GKE cluster - project-name - Name of the google project you wish to assign the GKE cluster Example: ./cluster.sh test-cluster-demo automation-test-demo

Options : build / delete / creds

build - builds your cluster with predefined settings delete - deletes your cluster creds - configures your kubectl config in order to use the cluster

Deploying the monitoring components

Once you you have finished building your cluster and gotten your credntials using the ./cluster.sh script. Simply run ./deploy.sh in order to deploy all the components

Accessing Grafana

Two things are needed to access grafana:

1. Credentials Username is admin Password: In order to get the password run kubectl get secret --namespace default grafana -o jsonpath="{.data.admin-password}" | base64 --decode ; echo in the terminal
2. This basic setup and configuration does not have an ingress so in order to access it from your machine you would have to expose the service: kubectl port-forward svc/grafana 8000:80 Once you have the port-forwarding in place you can navigate to http://localhost:8000 and login

Complete setup example

1. ./cluster.sh test-cluster-demo automation-test-demo - select build, wait for it to complete
2. ./cluster.sh test-cluster-demo automation-test-demo - select creds, wait for it to complete
3. ./deploy.sh - To build your resources
4. kubectl get secret --namespace default grafana -o jsonpath="{.data.admin-password}" | base64 --decode ; echo - to get login credentials
5. kubectl port-forward svc/grafana 8000:80 - To expose the service Note: If you are unable to expose the service due to pod being in a Pending state, please wait a few minutes and retry.
6. http://localhost:8000