jellyfin-ffmpegof

Unofficial docker image made by including ffmpegof in official Jellyfin docker image.

Note:

• This image uses /config/cache for cache dir by default, instead of the official /cache. This allows to use a single volume for stateful data, which can save costs when using Kubernetes in the cloud.
• The public ssh key is located inside the container at /config/ffmpegof/.ssh/id_ed25519.pub
• The known_hosts file is located inside the container at /config/ffmpegof/.ssh/known_hosts

Setup

Database

SQLite

SQLite is already provided and configured, you can start by adding workers.

Postgresql

If you want to use this container as a stateless app (currently not possible because Jellyfin itself isn't stateless) set the required fields in /config/ffmpegof/ffmpegof.yaml or equivalent env vars that take precedence for Postgresql:

Name	Default value	Description
FFMPEGOF_DATABASE_TYPE	sqlite	Must be 'sqlite' or 'postgres`
FFMPEGOF_DATABASE_HOST	localhost	Postgres database host
FFMPEGOF_DATABASE_PORT	5432	Postgres database port
FFMPEGOF_DATABASE_NAME	ffmpegof	Postgres database name
FFMPEGOF_DATABASE_USERNAME	postgres	Postgres database username
FFMPEGOF_DATABASE_PASSWORD	""	Postgres database password

Workers

Workers must have access to Jellyfin's /config/cache, /config/transcodes and /config/data/subtitles directories. It's recommended to setup NFS share for this.

For a worker docker image you can use this.

Adding new workers

Generate a new ssh key for Jellyfin:

docker compose exec -it jellyfin ssh-keygen -t ed25519 -f /config/ffmpegof/.ssh/id_ed25519 -q -N ""

Absolutely make sure that /config/ffmpegof/.ssh has 700 or 600 chmod-ed permissions, otherwise the remote ssh commands will fail with UNPROTECTED PRIVATE KEY FILE! error

Copy the public ssh key to the worker if it supports password login:

docker compose exec -it jellyfin ssh-copy-id -i /config/ffmpegof/.ssh/id_ed25519.pub root@<worker_ip_address>

If the worker doesn't support password login, you will have to copy the public key manually.

If you're using my rffmpeg-worker image, then the copied keys are stateless. Either add a volume to /etc/authorized_keys or make a file binding for the required ssh keys

Add the worker to ffmpegof db:

docker compose exec -it jellyfin ffmpegof add [--weight 1] [--name first_worker] <worker_ip_address>

Check the status of ffmpegof:

docker compose exec -it jellyfin ffmpegof status

Hardware Acceleration

Enable it normally in the Jellyfin admin panel.

If you want to use Hardware Acceleration all of the workers must support the same tech (VAAPI, NVENC, etc.).

Note: If the Jellyfin host doesn't support that same Hardware Accel tech then it can't be used as a failover, but if you have available workers it will still transcode without problems.

Kubernetes

On Kubernetes you can use OpenEBS to create RWX from RWO volume or Longhorn RWX volumes (NFSv4) and mount said paths to Jellyfin host and workers (must be exactly the same mount points!).

Here's a Helm chart repo with instuctions