Skip to content

Latest commit

 

History

History
233 lines (149 loc) · 8.44 KB

README.md

File metadata and controls

233 lines (149 loc) · 8.44 KB
Raw

Kubernetes Infrustructure Bootstrap

Base services collection for production ready Kubernetes cluster.

This collection based on know enterprise wide services, already used by many companies. Bootstrap contain base GitOps setup and examples for add new applications.

Target

Goal of this bootstrap define vendor free bootstrap, which can be started in any type of cluster, not depend on cloud provider.

But if it not problem for you, you can use Bitnami Kubernetes Production Runtime. BKPR alredy accessable for GKE, AKS, and Amazon EKS.

You can fork this project for define your own infrsutructure bootstrap. Any PRs are allways welcome.

Advice: Don't mix infrustructure and business services, this setup only define infrsutructure, for define business services better use GitOps soltuions, like ArgoCD, which already build it in this bootstrap.

Services

Current setup contains:

  • Kubernetes Dashboard - General-purpose web UI for Kubernetes clusters
  • Cert-Manager - Automatically provision and manage TLS certificates in Kubernetes
  • kube-prometheus-stack - kube-prometheus-stack 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.
  • loki-stack - Loki: like Prometheus, but for logs
  • tempo-distributed - Grafana Tempo in MicroService mode
  • argo-cd - Declarative continuous deployment for Kubernetes, GitOps implementation.

Requirements

Create kubernetes cluster version and configure kubectl for connect to it.

Install CLIs:

  • Helm - The package manager for Kubernetes.
  • Helm Diff - A helm plugin that shows a diff explaining what a helm upgrade would change
  • Hemlile - One file for manage multiple heml charts.
  • GNU Make - install by sudo apt-get install build-essential

First Start Guide

helmfile uses enviroment variables for set parametors of charts. Makefile automatically setup values from .env file.

  1. Copy .env.example and name it as .env
  2. Change variables in .env as you want.
  3. Run make setup - will upload all services without check on changes.

Usage

For setup basic infrustructure run

# Will deploy new or changed charts first time
make setup

For update only changed charts run

make sync

Setup certificates

For use https you need setup sertificates, you can do it by next commands

# Will be used by lets encrypt for send emails about certificate updates
export [email protected]
# Will create issuers (certificate providers)
make certificate-issuers

if make sync were made first time in cluster wait some time before setup certificates, k8s need time for load certificate manager operator

More about ceertificate configuration and tutorial for lets encrypt

IMPORTANT: If you not setup sertificates or setup them incorrectly, Ingresses will fallback to self-signed sertificates.

Minikube

For start local cluster and synchronise, just run

make local

It will run next commands, but you can run them by self:

# Start minikube server
minikube start

# Enable minikkube ingress before sync
make minikube-ingress # or minikube addons enable ingress

# for synchromise cluster
make setup # or helmfile sync

Get external ip, for access your cluster outside

make minikube-ip

And add to /etc/hosts file next line

# For access local cubernetes cluster
<your-external-ip> argo.k8s.local dashboard.k8s.local prometheus.k8s.local thanos-gateway.k8s.local grafana.k8s.local alertmanager.k8s.local k8s.local

Access Kubernetes Dashboard

For access kubernets dashboard you need firstly get token:

# list existing secrets
kubectl -n kubernetes-dashboard get secrets
# pass correct name of secret
kubectl -n kubernetes-dashboard describe secret kubernetes-dashboard-token-<some-id>
# copy token

Then you can open page and pass token

Though Ingress

open https://dashboard.k8s.local and pass copied token

Through proxy

create local proxy

kubectl proxy

open http://localhost:8001/api/v1/namespaces/kubernetes-dashboard/services/https:kubernetes-dashboard:https/proxy/ and pass copied token

Metrics, Logs, Tracing

In general this is self containing solution, which must just work out of the box.

For metrics collection used Prometheus. For togs collection Promtail and Loki. For traing collection used Tempo. For dashboard used Grafana.

In future release I would like to migrate all what possible to cloud IaaS solutions.

Why not ELK stack?

You can find comparisions from Grafana guys there. In simple words, Loki + Prometheus + Tempo + Grafana is simpler to setup then ELK, but it have some limitations. I actually love Kibana, and have plans to add it.

Why not OpenTelemetry?

I love OpenTelemetry idea of vendor agnostic fully containing stack, but it not ready for most of languages (in alpha or beta stages) right now. I would like to swith to OpenTelemetry when it will be ready for production.

Acording to their roadmap I've expecting to add OpenTelemetry in 2022, when they will add logs component support.

Acesss Grafana

You can open Grafana at grafana.k8s.local for login as admiin use username admin and password prom-operator

Change password in helfile.yaml in kube-prometheus-stack grafana section.

Access Logs

Open explore tab in Grafana, abd swith Prometheus to Loki. On log browser you can see posible valuues to search.

Access Traces

Open explore tab in Grafana, abd swith Prometheus to Tempo.

If Tempo not connected probably you need enable Tempo.

Enable Tempo

Open Configuration.DataSourses page in Grafana -> click Add data sourses -> click Tempo -> fill URL with http://tempo-tempo-distributed-query-frontend:3100 and set Trace to Logs section with Data Source Loki

Access ArgoCD

Get password to admin account

kubectl -n argocd get secret argocd-initial-admin-secret -o jsonpath="{.data.password}" | base64 -d && echo
# copy password

open https://argo.k8s.local and use admin as username

Create Application

You can easily [create application through:

Declarative application setup

For deploy applications from applications folder just run.

make apps # Will deploy application configurations

For add new applications just add new yaml in applications folder, like example application. You also can use this foler for setup Project, or repository, or app of apps.

Access deployed application

You can access already deployed application thourgh port-forwarding, just run:

kubectl port-forward service/frontend :80
# will choose local port and proxy it to service with name frontend (example application)

and you can open page localy as http://localhost:<allocated-port>