/cluster-monitoring

Cluster monitoring stack for clusters based on Prometheus Operator

Primary LanguageJsonnetMIT LicenseMIT

Cluster Monitoring stack for ARM / X86-64 platforms

The Prometheus Operator for Kubernetes provides easy monitoring definitions for Kubernetes services and deployment and management of Prometheus instances.

This have been tested on a hybrid ARM64 / X84-64 Kubernetes cluster deployed as this article.

This repository collects Kubernetes manifests, Grafana dashboards, and Prometheus rules combined with documentation and scripts to provide easy to operate end-to-end Kubernetes cluster monitoring with Prometheus using the Prometheus Operator. The container images support AMD64, ARM64, ARM and PPC64le architectures.

The content of this project is written in jsonnet and is an extension of the fantastic kube-prometheus project.

If you like this project and others I've been contributing and would like to support me, please check-out my Patreon page!

Components included in this package:

  • The Prometheus Operator
  • Highly available Prometheus
  • Highly available Alertmanager
  • Prometheus node-exporter
  • kube-state-metrics
  • CoreDNS
  • Grafana
  • SMTP relay to Gmail for Grafana notifications (optional)

There are additional modules (disabled by default) to monitor other components of the infra-structure. These can be enabled or disabled on vars.jsonnet file by setting the module enabled flag in modules to true or false.

The additional modules are:

  • ARM-exporter to generate temperature metrics (works on some ARM boards like RaspberryPi)
  • MetalLB metrics
  • Traefik metrics
  • ElasticSearch metrics
  • APC UPS metrics
  • GMail SMTP relay module

There are also options to set the ingress domain suffix and enable persistence for Grafana and Prometheus.

The ingresses can use TLS with the default self-signed certificate from your Ingress controller by setting TLSingress to true and use a custom certificate by creating the files server.crt and server.key and enabling the UseProvidedCerts parameter at vars.jsonnet.

Persistence for Prometheus and Grafana can be enabled in the enablePersistence section. Setting each to true, creates the volume PVCs. If no PV names are defined in prometheusPV and grafanaPV, the default StorageClass will be used to dynamically create the PVs The sizes can be adjusted in prometheusSizePV and grafanaSizePV.

If using pre-created persistent volumes (samples in samples), check permissions on the directories hosting the files. The UID:GID for Prometheus is 1000:0 and for Grafana is 472:472.

Changing these parameters require a rebuild of the manifests with make followed by make deploy. To avoid installing all pre-requisites like Golang, Jsonnet, Jsonnet-bundler, use the target make docker to build in a container.

Quickstart (non K3s)

The repository already provides a set of compiled manifests to be applied into the cluster or the deployment can be customized thru the jsonnet files.

If you only need the default features and adjust your cluster URL for the ingress, there is no need to rebuild the manifests(and install all tools). Use the change_suffix target with argument suffix=[suffixURL] with the URL of your cluster ingress controller. If you have a local cluster, use the nip.io domain resolver passing your_cluster_ip.nip.io to the suffix argument. After this, just run make deploy.

# Update the ingress URLs
make change_suffix suffix=[suffixURL]

# Deploy
make deploy

To customize the manifests, edit vars.jsonnet and rebuild the manifests.

$ make vendor
$ make
$ make deploy

# Or manually:

$ make vendor
$ make
$ kubectl apply -f manifests/setup/
$ kubectl apply -f manifests/

If you get an error from applying the manifests, run the make deploy or kubectl apply -f manifests/ again. Sometimes the resources required to apply the CRDs are not deployed yet.

If you enable the SMTP relay for Gmail in vars.jsonnet, the pod will be in an error state after deployed since it would not find the user and password on the "smtp-account" secret. To generate, run the scripts/create_gmail_auth.sh script.

Quickstart on Minikube

You can also test and develop the monitoring stack on Minikube. First install minikube by following the instructions here for your platform. Then, follow the instructions similar to the non-K3s deployment:

# Start minikube (if not started)
minikube start

# Enable minikube ingress to allow access to the web interfaces
minikube addons enable ingress

# Get the minikube instance IP
minikube ip

# Run the change_suffix target
make change_suffix suffix=[minikubeIP.nip.io]

# or customize additional params on vars.jsonnet and rebuild
make vendor
make

# and deploy the manifests
make deploy

# Get the URLs for the exposed applications and open in your browser
kubectl get ingress -n monitoring

Quickstart for K3s

To deploy the monitoring stack on your K3s cluster, there are four parameters that need to be configured in the vars.jsonnet file:

  1. Set k3s.enabled to true.
  2. Change your K3s master node IP(your VM or host IP) on k3s.master_ip parameter.
  3. Edit suffixDomain to have your node IP with the .nip.io suffix or your cluster URL. This will be your ingress URL suffix.
  4. Set traefikExporter enabled parameter to true to collect Traefik metrics and deploy dashboard.

After changing these values to deploy the stack, run:

$ make vendor
$ make
$ make deploy

# Or manually:

$ make vendor
$ make
$ kubectl apply -f manifests/setup/
$ kubectl apply -f manifests/

If you get an error from applying the manifests, run the make deploy or kubectl apply -f manifests/ again. Sometimes the resources required to apply the CRDs are not deployed yet.

If you enable the SMTP relay for Gmail in vars.jsonnet, the pod will be in an error state after deployed since it would not find the user and password on the "smtp-account" secret. To generate, run the scripts/create_gmail_auth.sh script.

Ingress

Now you can open the applications:

To list the created ingresses, run kubectl get ingress --all-namespaces, if you added your cluster IP or URL suffix in vars.jsonnet before rebuilding the manifests, the applications will be exposed on:

Updating the ingress suffixes

To avoid rebuilding all manifests, there is a make target to update the Ingress URL suffix to a different suffix. Run make change_suffix suffix="[clusterURL]" to change the ingress route IP for Grafana, Prometheus and Alertmanager and reapply the manifests.

Customizing

The content of this project consists of a set of jsonnet files making up a library to be consumed.

Pre-reqs

The project requires json-bundler and the jsonnet compiler. The Makefile does the heavy-lifting of installing them. You need Go already installed:

git clone https://github.com/carlosedp/cluster-monitoring
cd cluster-monitoring
make vendor
# Change the jsonnet files...
make

After this, a new customized set of manifests is built into the manifests dir. To apply to your cluster, run:

make deploy

To uninstall, run:

make teardown

Images

This project depends on the following images (all supports ARM, ARM64 and AMD64 thru manifests):

Alertmanager Blackbox_exporter Node_exporter Snmp_exporter Prometheus

ARM_exporter

Prometheus-operator

Prometheus-adapter

Grafana

Kube-state-metrics

Addon-resizer

Obs. This image is a clone of AMD64, ARM64 and ARM with a manifest. It's cloned and generated by the build_images.sh script

configmap_reload

prometheus-config-reloader

SMTP-server

Kube-rbac-proxy