Generate a complete Helm chart — Chart.yaml, values, helpers, Deployment, Service and optional Ingress, HPA and test — with the label and name rules handled.
my-app · 8 files
Chart version 0.1.0, app version 1.0.0. Nothing has been validated by Helm.
The chart's identity. `version` is the chart's own SemVer; `appVersion` is the application's and need not be SemVer.
apiVersion: v2
name: my-app
description: A Helm chart for my-app
type: application
# The chart's own version. Bump it on every change to the chart.
version: 0.1.0
# The version of the application this chart deploys. Quoted so a version
# like 1.10 is not read as the number 1.1.
appVersion: "1.0.0"
These are Go templates with Sprig functions, and nothing here executes Go templates — so no file has been rendered, checked by helm lint, or validated against a cluster. Run helm template . to see what they actually produce and helm lint to check them against the schema. One thing the generated chart does for you: the Deployment’s replicas are guarded when an HPA is included, so an upgrade cannot overwrite the replica count the autoscaler just chose.
Helm Chart Generator runs entirely on your device — no Helm command is run and nothing is installed.
You’re on the Free plan. Compare plans
The `_helpers.tpl` is where charts go wrong, so it is the part written most carefully. Generated names are piped through `trunc 63 | trimSuffix "-"`, because 63 is the limit for a label value and truncating at 63 can leave a trailing dash, which is itself an invalid name. Two separate problems that look like one.
The selector labels are a deliberate subset of the full labels. A Deployment's selector is immutable after creation, so it must not include the version label that changes on every release — a chart that puts it there cannot be upgraded without deleting the Deployment.
When an HPA is included, the Deployment's replicas are guarded behind `if not .Values.autoscaling.enabled`. Without that, every `helm upgrade` writes the static replica count back over whatever the autoscaler just decided, so the deployment scales down to the chart default in the middle of the traffic that scaled it up. It is one `if` and it is the difference between an HPA working and an HPA fighting your pipeline.
The image tag is left empty on purpose, with the template defaulting to `.Chart.AppVersion`. That keeps the deployed image and the chart's declared app version in step automatically, which is the convention and the reason `appVersion` exists at all.
Nothing is rendered. These are Go templates with Sprig functions, and this does not execute Go templates — so no file here has been checked by Helm. `helm template .` shows what they produce and `helm lint` checks them against the schema.
So the template can fall back to `.Chart.AppVersion`, which keeps the deployed image and the chart's declared app version in step. Bump appVersion and the image follows; hardcode the tag and you have two places to remember.
Label values are limited to 63 characters, so a generated name has to be truncated. Truncating can land on a dash, and a name ending in a dash is invalid — so the trim is not decoration, it is the second half of the fix.
Because the Deployment's replica count and the HPA both own that field. Without the guard, `helm upgrade` writes the chart's static default back over the autoscaler's decision, scaling you down during exactly the traffic that scaled you up.
It is a complete and correctly shaped chart, but nothing here has rendered or validated it — these are Go templates and this does not execute them. Run `helm lint` and `helm template .` before installing anything.