finds.dev← search

// the find

8gears/n8n-helm-chart

★ 705 · Go Template · Apache-2.0 · updated Sep 2026

A Kubernetes Helm chart for n8n - a workflow automation tool. Easily automate tasks across different services on self hosted onKubernetes

A community-maintained Helm chart for running n8n on Kubernetes, modeling the main/worker/webhook split that n8n's queue-mode architecture actually uses rather than treating it as a single stateless deployment. Aimed at teams self-hosting n8n at a scale beyond a single Docker container who need Postgres, Redis/Valkey, and optionally the AI sandbox wired up declaratively.

Probes are split correctly across startup/liveness/readiness, using /healthz vs /healthz/readiness so a DB outage drains pods from the Service instead of restarting all of them at once — and the values.yaml comments explain why, which most charts don't bother doing. The worker and webhook deployments are genuinely separate from main with their own scaling, resources, and persistence, matching n8n's real queue-mode topology instead of faking it with env var toggles on one deployment. config/secret dictionaries map 1:1 to env vars, so any new n8n setting works immediately without waiting on a chart update. The sandbox section documents real operational landmines (DB-stored AI settings silently overriding env vars, sysbox vs privileged tradeoffs, checking user.max_user_namespaces before enabling) that most people would only discover by debugging a stuck pod.

It's a third-party chart (8gears, not n8n-io) actively asking for more maintainers in the README banner, so there's real bus-factor risk if you depend on it long-term. No default resource requests/limits anywhere — a deliberate minikube-friendly choice, but it means a naive first install can OOM without warning. The project's own docs admit queue-mode HA is partial: the main instance stays a single point of failure, which will surprise anyone assuming 'queue mode' means true high availability. The sandbox feature's most isolated runner mode needs an n8n Enterprise license plus mutual TLS with a private CA and sysbox node prep — a lot of cluster surgery for something gated behind an 'enabled: false' flag.

View on GitHub → Homepage ↗

// want more like this?

We dig through GitHub every week and send a few repos picked for what you actually care about — each with an honest take like this one.

Get finds in your inbox → Search again →