Dozzle is a lightweight web interface for watching Docker container logs in real time. It is a single small container that mounts the Docker socket and serves a live view of every container’s stdout and stderr in your browser — no Elasticsearch, no database, no log shipping pipeline. Nothing is written to disk: Dozzle is built for reading what is happening right now, not for archiving it.
That focus is the point. The image is around 7 MB compressed, memory use is in the tens of megabytes, and starting it is one docker run with a socket mount. Behind the simple premise sits a competent log tool: regex and SQL-style filtering, fuzzy container search, split-screen streams, live CPU and memory counters per container, and multi-user authentication for teams that cannot leave logs open to anyone who can reach the port.
/data volume.Dozzle talks to the Docker API directly over the mounted socket and streams container output to the browser over its own connection, so logs are never buffered to disk. API version negotiation is automatic, which is why it copes with a range of engines; the hard floor is Docker Engine 19.03 (API 1.40) or newer. On Podman, the remote socket has to be enabled first, because Podman is daemonless and has no engine identity to report — Dozzle derives a stable host id itself.
Configuration follows the twelve-factor convention: CLI flags or environment variables, documented in full on the project’s site. Dozzle collects anonymous usage configuration through Google Analytics, which the project explains as the only feasible way for an unfunded open-source tool to know how it is used; the collected data is public in a live dashboard.
The container is published as amir20/dozzle and the documented run is a single command that mounts the Docker socket and a named volume for /data, exposed on port 8080; a compose file with the same mounts is provided for anything longer-lived. The /data volume matters: user accounts and settings live there, so without it the login configuration is lost the next time the container is recreated. For remote hosts, the same image runs in agent mode on its own port. There is no package in any distribution repository — Docker or Podman is the delivery mechanism.
docker logs -f in a terminal per service.