Podos A lightweight container platform Early development · not released

Comparison

Comparison

Podos sits between running containers by hand and running a cluster orchestrator. This page compares it with tools people often consider for the same job. Descriptions of other projects are summaries; check their own documentation for details.

At a glance

Podos Kubernetes Kamal Coolify
What you install on the server one binary, crun, conmon, nft a distribution with control plane, container runtime, network plugin, ingress controller Docker Docker and Coolify's own containers
Container engine crun directly containerd or CRI-O Docker Docker
Control from CLI, and a web UI for viewing and for variables and secrets kubectl, API CLI on your machine, over SSH web UI
Application description one short manifest, TOML or YAML many YAML objects (Deployment, Service, Ingress, …) config/deploy.yml settings in the UI
State SQLite inside the daemon etcd none on the server beyond Docker its own database
Scope today one node (several planned for 0.4) many nodes several servers, each driven over SSH one or more servers

Kubernetes

Kubernetes is the standard for running containers on many machines. It is powerful and extensible, and it brings many moving parts: a control plane, a separate network plugin, an ingress controller, a certificate manager, and a large vocabulary of objects.

Podos takes a few ideas from that world (declarative desired state, namespaces, rolling updates with health checks) and leaves out the rest. One binary is the daemon, the API and the CLI. Networks, ingress, certificates and secrets are built in, with secure defaults, rather than chosen and wired together. Config files and key files, the job of a ConfigMap or a mounted Secret, are built in too (Available): podos file set and files in the manifest. Each namespace and environment gets its own network and cannot reach the others (Available), which Kubernetes does not do without extra network policy.

Choose Kubernetes when you need many nodes now, its ecosystem, or a platform team already runs it. Choose Podos when one server (or a few, later) is enough and you want to understand all of it.

Kamal

Kamal deploys Docker containers from your machine to your servers over SSH, with a proxy in front for zero-downtime switches and Let's Encrypt certificates. There is no daemon of its own on the server; Docker and SSH do the work.

Podos runs a daemon on the server instead. That costs one long-running process, and gives a place for state: the daemon knows what should run, keeps releases, and (Available) restarts containers and recovers them after a reboot without anyone running a command. Podos does not need Docker on the server, builds on the server (Available) instead of needing a registry, and stores secrets encrypted on the server per environment.

Coolify

Coolify is a self-hosted platform with a web interface, in the spirit of Heroku or Vercel. It runs on Docker and brings a proxy, its own database, and many one-click services.

Podos aims at fewer moving parts and a manifest in your repository as the source of truth, driven from the command line. The web UI is a view on the same API: it shows the server and lets you change variables and secrets, while deploys stay in the CLI. Logs and metrics screens are planned for 0.2 and an app catalog for 0.3.

When Podos is not the right choice