docker-job

docker-job

Run one-shot services on a Docker Swarm.

When using a Docker Swarm, if a container is launched with docker run it is not actually run by the swarm--it is run as a standalone container. This means that the lifecycle of the container's execution is not managed by the orchestrator or scheduler.

The docker-job command-line application launches an image by wrapping it in a service and running it. Options are available to determine whether to output the logs when the job has completed, run more than one replica at the same time, and so on.

Installation

Ensure you actually have a Docker Swarm, as per Run Docker Engine in swarm mode.

Install docker-job from NPM with:

npm i -g docker-job

Usage

Bare minimum is to provide the name of an image, which will be launched on your local Docker Swarm:

dj hello-world

Any parameters after the image name will be passed to the launched job:

dj alpine echo hello world

The app will provide an identifier for the service, and then wait until it completes before exiting. To show the logs on exit use the --showlogs option:

dj --showlogs hello-world

Alternatively, to exit immediately without waiting for completion, use the detach option (--detach or -d):

dj --detach hello-world

To force the image to be pulled from a registry before running, add the --pull option:

dj --pull hello-world

To launch a number of instances of the image, add the --replicas option:

dj --replicas 5 hello-world

Again, the --showlogs option can be used to show the logs at the end, as each instance finishes. (This is different to the way docker service logs behaves, where the logs from each instance are interleaved by time.)

To run the job repeatedly until some condition is met, use --repeat-until with a regular expression:

dj --showlogs --repeat-until '20:0[05]' alpine date

When the job has finished running the logs are checked and the presence of '20:00' or '20:05' anywhere in the logs will prevent docker-job from running the job again.

The logs can be shown on each iteration, and more than one replica can be run:

dj --showlogs --replicas 5 --repeat-until 'workflows?' hello-world

The sequence will terminate if any of the replicas outputs a string that matches the regular expression provided.

To repeat a job as long as some string exists in the logs, then use the --repeat-while option. The following example will run as long as the time is 8pm and any number of seconds:

dj --repeat-while '20:00:\d\d' alpine date

When a job has finished the service that supports it can be removed automatically with the --rm option:

dj --rm --repeat-while '20:00:\d\d' alpine date

To map a volume use --volume or -v:

dj -v /var/run/docker.sock:/var/run/docker.sock \
  tmaier/docker-compose docker-compose up

Alongside volumes Docker Swarm also supports config files, which can be accessed with --config:

dj \
  --showlogs \
  --config src=foo.sh,target=/usr/src/app/bar.sh \
  alpine \
    cat /usr/src/app/bar.sh

The source can also be expressed with 'source':

dj \
  --showlogs \
  --config source=foo.sh,target=/usr/src/app/bar.sh \
  alpine \
    cat /usr/src/app/bar.sh

and if no name/value pairs are present the entire string is treated as the source:

dj \
  --showlogs \
  --config foo.sh \
  alpine \
    cat /foo.sh

To provide one or more environment variables use --env or -e:

dj \
  -v /var/run/docker.sock:/var/run/docker.sock \
  --env DBHOST=server \
  --env DBPASS=xyz \
  tmaier/docker-compose \
    docker-compose up

To use a different host, set -H or --host:

dj -H http://swarm:2375 --rm hello-world

Note that for the moment the value must include the protocol and port.

Swarm on AWS

To use a swarm that is running on AWS, create an SSH tunnel with the ---ssh-* options. For example:

dj \
  --ssh-remote /var/run/docker.sock \
  --ssh-identify-file ./docker-swarm.pem \
  --ssh-hostname [e-mail] \
  hello-world

Note that there is an issue with logs not being available when using AWS.