Direct naar content

How to Postgres on Kubernetes

Several years ago now, colleague Craig wrote a three-part blog about PostgreSQL in combination with Docker. Technology has not stood still, so an update is in order, thought Nathan Koopmans, cloud platform engineer at OptimaData. Hence, How to Postgres on Kubernetes.

How to Postgres on Kubernetes

Differences between Docker and Kubernetes

In this blog, I will assume that you already have a working Kubernetes setup. Whether this is with MiniKube, K3s, a local complete Kubernetes setup or at a cloud provider like AWS, GCP or Azure, it doesn’t matter. Personally, I use our local Kubernetes environment. Kubernetes has a few advantages over Docker, including the self-healing bit. To make the differences between Docker and Kubernetes clear, I borrowed a nice image from the Internet.

Differences between Docker and Kubernetes

The basics

Let’s start at the beginning. Therefore, in part one I will show a setup based on a standard PostgreSQL image, in blog part two (in 2 weeks) I will use a Kubernetes database operator (CloudnativePG). But now first: Plain PostgreSQL.

First you create a YAML file to define in it some things for the PostgreSQL pod / container. What things are needed and what are they for?


A secret is a clean way to store data such as passwords. In the case of PostgreSQL, you specify a user and password to use to connect to the database. Secrets can be used for multiple pods and contain only a small amount of data. For all the ins and outs about secrets, visit the Kubernetes page on secrets.

apiVersion: v1

kind: Secret


  name: postgres-env

type: Opaque


  POSTGRES_USER: example




A PersistentVolumeClaim is a claim on a bit of space to use. It is attached to a pod so that data can be stored here. This has the advantage that if a pod crashes, the data is not lost, but taken into a new pod. You can find more info about PVCs and all the options available on the Kubernetes-website.

apiVersion: v1

kind: PersistentVolumeClaim


  name: postgres-pv-claim



  - ReadWriteOnce



      storage: 1Gi



The deployment takes care of creating and managing the pods within the deployment. When you scale the deployment from one to two pods, the deployment causes the cube scheduler to create the second pod. If you modify something in the deployment, the pods will be replaced one by one with the desired modifications. In the deployment you also specify which volumes (claims) are to be used. If desired, you can also define other resources, such as CPU and RAM usage. More info can be found on the Kubernetes website.

apiVersion: apps/v1

kind: Deployment



    app: postgres

  name: postgresql


  replicas: 1


    type: Recreate



      app: postgres




        app: postgres



      - envFrom:

        - secretRef:

            name: postgres-env


        name: postgresql


        - containerPort: 5432

          name: postgresql


        - mountPath: /var/lib/postgresql

          name: postgres-data


      - name: postgres-data


          claimName: postgres-pv-claim

In my case, it ends up looking like this:

Demo how to Postgres on Kubernetes

Demo 2 How to Postgres on Kubernetes

Note the “---” as I used it in the screenshot above. Don’t forget to add it!

Best practise

To keep everything clean and organized, create a separate namespace where you deploy this container. Since I don’t have a separate namespace for this blog yet, I’ll create it first:

kubectl create ns postgresdeploy

namespace best practise for Postgres on Kubernetes deployment

The namespace has been successfully created. Now it is time to apply the YAML file so that the Secret, PersistentVolumeClaim and Deployment are created:

kubectl apply -f postgresdeploy.yaml -n postgresdeploy

YAML-file in Postgres on Kubernetes deploymentDon’t you get an error message? Then you did it right. If you do get an error message, it is usually due to a typo or indentation. YAML is very sensitive to the alignment of the various components in the file.


Give the process a few minutes to create and start everything. After that you can check if everything is created and the container is ready and has as Running status.

kubectl get all -n postgresdeploy

Running check how to Postgres on Kubernetes deployment

What is noticeable is that the PVC is missing here. This is correct because a PVC is cluster-wide. Want to check if the PVC has been created? Then run the following command:

kubectl get pvc -n postgresdeploy

PVC check how to Postgres on Kubernetes deployment


Now you are going to connect to the container:

kubectl exec -it postgresql-d684bfd65-5cjzm -n postgresdeploy -- /bin/bash

You will then get the bash prompt. Now enter:

su - postgres

This will make you become the postgres user that is active within the postgres image by default. Next, type the following:

psql -U example

And voila, you are in postgres. There is no need to enter a password this way. Connecting externally does ask for the password. I then use \l+ to check that I am getting data back and everything is responding as it should.

Postgres on Kubernetes deployment check