Skip to main content
Helm is a package manager for Kubernetes. It is to a cluster roughly what Docker Compose is to a single machine: one command installs the app and the things it needs. This page is for people who already operate a Kubernetes cluster. If you do not, you almost certainly want Docker Compose instead, which is the supported path and considerably less work.

​
Before you start

  • A Kubernetes cluster and kubectl pointed at it
  • Helm 3
  • An ingress controller, if you want her reachable from outside the cluster
  • A Temporal server, which the chart does not provide
The chart does not bundle Temporal. PostgreSQL and Redis come with it, Temporal does not, and Temporal is what actually fires your scheduled posts. Without a reachable one the interface loads, you can write posts, and nothing is ever published. Install Temporal on your cluster or use Temporal Cloud, then point TEMPORAL_ADDRESS at it.

​
Install

The chart is published as an OCI artifact, so there is no helm repo add step:
The chart is called postqueen-app, not postqueen. The directory in the repository is charts/postqueen, but the name inside Chart.yaml, which is what the registry uses, is postqueen-app. The directory name will not resolve as an OCI reference.
If you use Flux or Argo CD, point them at the same reference in OCI mode.

​
A minimal values.yaml

Nothing derives the connection strings for you, not even for the bundled databases, so they have to be written out. The hostnames below assume a release named postqueen, which is what the command above creates.
The credentials inside those two connection strings are the chart’s own defaults for the bundled PostgreSQL and Redis. If you change them under postgresql.auth or redis.auth, change them here to match. To use databases you already run, point both strings at those and set postgresql.enabled: false and redis.enabled: false. Every variable is explained in Configuration Reference. Values under secrets become a Kubernetes Secret, values under env become a ConfigMap, and both are handed to the container.

​
Traps worth knowing about

Published tags look like v3.0.4. When image.tag is empty the chart falls back to its appVersion, which carries no v, so it resolves to an image that does not exist and the pod sits in ImagePullBackOff. Write the tag out, with the v.
With no extraVolumes configured, /uploads is an emptyDir, which lives and dies with the pod. For a real install, either mount a persistent volume through extraVolumes and extraVolumeMounts, or use Cloudflare R2. R2 is also the answer for more than one replica, since a local disk cannot be shared between pods.
The deployment defines no liveness or readiness probes, so traffic reaches a pod that is still starting, and a wedged pod is never restarted. Add your own against / on the container port if you want that behaviour.
Every key under secrets is rendered whether or not you gave it a value, so an unset provider key arrives as an empty string rather than being absent. Harmless in practice, and useful to know when something looks set but is not.
A release named postqueen produces objects named postqueen-postqueen-app. That is the standard Helm naming helper, not a mistake. The bundled databases are unaffected and are named postqueen-postgresql and postqueen-redis-master, which is why the connection strings above look the way they do.

​
Check it worked

You want Backend started successfully on port 3000, and then the lines after it. She reports configuration problems there and carries on running regardless, so a Running pod is not proof that the settings are right:
The chart lives at GkhanKINAY/postqueen-helmchart, and the full list of values is in charts/postqueen/values.yaml.

​
Next Steps