Wat?

If on Windows, set the environment variable: COMPOSE_CONVERT_WINDOWS_PATHS=1 so /var/run/docker.sock works.
docker-compose up (specify optional override files with different volume mappings)
optionally --scale process=<n> where <n> is a sensible number (5 seems ok)
optionally --no-start so it creates the containers but you can up the stack manually
Ensure the logging stack is up before initiating processing!
- logspout will crash out because logstash isn't up
- sometimes elasticsearch needs some hand holding (restarting) to come up ok, not sure why...
- once es and kibana are up, logstash will finally roll in
- once logstash is listening, restart logspout
POST to localhost:5000/initiate to trigger an enumeration of files in the folder you mount to /var/work (see the compose file)

That will start everything running.

Redis, ElasticSearch and Kibana are exposed so you can try and see what's going on. Check the compose file for ports.

Process

bootstrap literally just listens for a POST so it can tell enqueue to start things off. This allows you to populate your data drop first. We aren't cool enough for folder watching.
enqueue polls redis every n seconds and when bootstrap queues a message it enumerates the work directory, one file at a time, queues the filename in redis and then copies the file to the queued subdirectory.
process polls redis every n seconds and when it sees a filename, processes that filename (if it has an accepted extension e.g. .csv), posting row by row to Elastic Search.
mq-api is a REST API for the redis message queue library (rsmq). the dotnet app hits this to interact with the queue (the node apps use the library natively)
the rest of the stack is standard images