Simple scheduling for short-running Docker containers
Dockron requires access to the Docker, so it may need to be run as root, or, if in a Docker container, need the socket mapped as a volume.
As simple as:
dockron
It will then run in the foreground, periodically checking Docker for containers with labels containing a cron schedule.
By default, Dockron will periodically poll Docker for new containers or schedule changes every minute. You can specify an interval by using the -watch flag.
Dockron is also available as a Docker image. The multi-arch repo can be found at IamTheFij/dockron
From either an amd64, arm, or arm64 machine, you can run Dockron using:
docker run -v /var/run/docker.sock:/var/run/docker.sock:ro iamthefij/dockron -watch
You might see something like the following error when Dockron connects to the Docker API
Error response from daemon: client version 1.47 is too new. Maximum supported API version is 1.45
This is because the API client library is newer than the version of the Docker API on your host. You can tell the Dockron API Client to use a compatible version by specifying DOCKER_API_VERSION=1.45, where the version you specify matches the API version shown when you run docker version. If you are running Dockron in Docker, make sure you add this to your compose environment or otherwise pass it to the container.
First, be sure your container is something that is not long running and will actually exit when complete. This is for batch runs and not keeping a service running. Docker should be able to do that on it's own with a restart policy.
Create your container and add a label in the form 'dockron.schedule=* * * * *', where the value is a valid cron expression (See the section Cron Expression Formatting).
Dockron will now start that container peridically on the schedule.
If you have a long running container that you'd like to schedule an exec command inside of, you can do so with labels as well. Add your job in the form dockron.<job>.schedule=* * * * * and dockeron.<job>.command=echo hello. Both labels are required to create an exec job.
Eg.
labels:
- "dockron.dates.schedule=* * * * *"
- "dockron.dates.command=date"
Note: Exec jobs will log their output to Dockron. There is also currently no way to health check these.
For more information on the cron expression parsing, see the docs for robfig/cron.
Dockron is quite simple right now. It does not yet:
- Issue any retries
- Cancel hanging jobs
I intend to keep it simple as well. It will likely never:
- Provide any kind of alerting (check out Minitor)
- Handle job dependencies
Either use a separate tool in conjunction with Dockron, or use a more robust scheduler like Tron, or Chronos.
If you have go on your machine, you can simply use make build or make run to build and test Dockron. If you don't have go but you do have Docker, you can still build docker images using the provide multi-stage Dockerfile! You can kick that off with make docker-staged-build
There is also an example docker-compose.yml that will use the multi-stage build to ensure an easy sample. This can be run with make docker-example.
There are now some basic tests as well as linting and integration tests. You can run all of these by executing make all.