# OpenFaaS — Serverless Functions Made Simple

> OpenFaaS is a framework for packaging code, binaries or containers as portable serverless functions that run on Kubernetes or containerd with autoscaling, metrics, and a rich CLI.

## Install

Save in your project root:

# OpenFaaS — Serverless Functions Made Simple

## Quick Use
```bash
# Install on Kubernetes with arkade
arkade install openfaas
# Forward the gateway and log in
kubectl -n openfaas port-forward svc/gateway 8080 &
export OPENFAAS_URL=http://127.0.0.1:8080
faas-cli login --username admin --password-stdin < password.txt

# Scaffold and deploy a Python function
faas-cli new hello --lang python3-http
faas-cli up -f hello.yml
curl $OPENFAAS_URL/function/hello -d ''hi''
```

## Introduction
OpenFaaS turns any container into a function with a uniform HTTP interface and scale-to-zero behavior. It is the most popular open-source FaaS platform and runs on anything that speaks the Kubernetes or containerd APIs — including Raspberry Pi clusters.

## What OpenFaaS Does
- Packages handlers as OCI images with a minimal watchdog process
- Exposes functions behind a single gateway with auth and metrics
- Autoscales via Prometheus alerts or HPA v2
- Supports async invocations through NATS JetStream
- Ships templates for Python, Node, Go, Ruby, Rust, PHP, Java and more

## Architecture Overview
A gateway service (Go) routes HTTP requests to per-function Deployments; each function pod runs a watchdog that forks the handler or proxies HTTP. Metrics flow to Prometheus, which triggers scaling decisions. An optional queue-worker pulls async jobs from NATS and invokes functions with retries.

## Self-Hosting & Configuration
- `arkade install openfaas` or the `openfaas` Helm chart
- faasd variant runs on a single VM via containerd (no K8s needed)
- Use `OPENFAAS_URL` and basic-auth/OIDC for the gateway
- Define functions in a `stack.yml` — `faas-cli up` builds, pushes, deploys
- Wire Prometheus + Grafana dashboards shipped with the chart

## Key Features
- True scale-to-zero with configurable idle timeout
- Async invocations with NATS-backed retries and callbacks
- Function templates store and `of-watchdog` streaming handlers
- Works on ARM64 — great for edge and homelab
- Pro version adds multi-tenancy, CRDs and single sign-on

## Comparison with Similar Tools
- **Knative Serving** — heavier, tied to Kubernetes and Istio/Contour
- **Fission** — similar idea, pool-based cold starts, less active
- **AWS Lambda** — managed only, provider lock-in
- **Kubeless** — archived
- **Nuclio** — data-science focused, richer triggers

## FAQ
**Q:** Do I need Kubernetes?
A: No — `faasd` runs the core stack on one VM with systemd + containerd.

**Q:** How do cold starts compare to Lambda?
A: Warm pods are sub-10 ms; cold starts depend on image pull and usually land at 1–3 s.

**Q:** Can I use my own container image?
A: Yes — set `image:` and `fprocess:` in `stack.yml` and skip the template.

**Q:** Is OpenFaaS CE still open source?
A: Yes, MIT licensed; OpenFaaS Pro adds commercial features under EULA.

## Sources
- https://github.com/openfaas/faas
- https://docs.openfaas.com/

---
Source: https://tokrepo.com/en/workflows/258f31fe-38fa-11f1-9bc6-00163e2b0d79
Author: AI Open Source