The Kubernetes dispatcher is built and designed to be coupled with the dispatcher operator. This code interacts with the dispatcher library to download data, route events, call a sidecar container, and upload results all within Kubernetes.
There are several parts to this code as it encompasses integrating several python libraries together.
For each major library we have integration points in specific modules to handle configuration of each library.
The configuration of PeeWee is pulled from an INI file parsed from an environment variable or command line option. The configuration in the file is a database connection url.
The CherryPy configuration has two entrypoints for use. The WSGI interface and the embedded server through the main method.
The Celery tasks are located in their own module and have an entrypoint from the CherryPy REST objects. The tasks save state into a PeeWee database that is also accessed in the CherryPy REST objects.
The default way to start up this service is with a shared SQLite database. The database must be located in the current working directory of both the celery workers and the CherryPy web server. The messaging system in Travis and Appveyor is redis, however the default is rabbitmq.
There are three commands needed to start up the services. Perform these steps in three separate terminals.
docker-compose up redis postgrescelery -A pacifica.dispatcher_k8s.tasks worker -l infopython -m pacifica.dispatcher_k8s
To test working system run the following in bash:
UUID=$(curl http://127.0.0.1:8069/dispatch/add/2/2)curl http://127.0.0.1:8069/status/$UUID