Grow within your enemy, then burst out.

Discouraged, google has since enabled python as a proper function engine, and this project now only serves as insight into the inner workings of google cloud functions.

• Docker
• Service Account json with Google Cloud Functions Developer role
• Pipenv
• Project with a flask blueprint and an entry point called cloud.py

import flask
import cloud_functions

flask_app = flask.Flask(__name__)


@flask_app.route('/')
def hello():
    return "Hello cloud functions."


cloud_functions.register_http_trigger(flask_app)

FROM stormydragon/gcf-python
CMD ["--http", "--project=<my project name>", "--name=<trigger name>"]

pipenv install flask
docker build --tag my_cloud_function .
docker run --rm -it -v /path/to/service-account.json:/service-account.json:ro my_cloud_function

That's good too, simply pick a resource and event type from below and use code similar to this you can find examples of the context and data in the invocation data.md file.

cloud.py

import cloud_functions
import logging
logger = logging.getLogger(__name__)


def handler(context, data):
    logger.info(context, data)


cloud_functions.register_event_trigger(handler)

Dockerfile

FROM stormydragon/gcf-python
CMD ["--event", "--project=<my project name>", "--name=<trigger name>", "--resource=<the resource>", "--event=<the event>"]

pipenv install flask
docker build --tag my_cloud_function .
docker run --rm -it -v /path/to/service-account.json:/service-account.json:ro my_cloud_function

The container will package your project during build and by running it with the specified arguments will deploy the package to Google Cloud Functions

Replace google cloud function node interpreter with arbitrary code Instead of using node as a shim, simply replace node.js and be master control program of the container.

The node worker script google_cloud_worker/worker.js was built into the container by the cloud function deployment machinery and is untouchable by our code until the /load endpoint is called, this is triggered by a "cold start" before the first connection is made. Our code usurps the node process and executes our own, in the process handing the connected /load and listening file descriptor to it, it is then that we send the answer to load, and accept new connections on the listener.

Google cloud functions consists of a webserver that accepts three 'paths' as well as communication via http with a supervisor service.

• X_GOOGLE_CODE_LOCATION
• X_GOOGLE_ENTRY_POINT
• X_GOOGLE_FUNCTION_TRIGGER_TYPE - HTTP_TRIGGER, ,
• X_GOOGLE_FUNCTION_NAME
• X_GOOGLE_FUNCTION_TIMEOUT_SEC
• X_GOOGLE_WORKER_PORT - Web server port
• X_GOOGLE_SUPERVISOR_HOSTNAME
• X_GOOGLE_SUPERVISOR_INTERNAL_PORT

• MAX_LOG_LENGTH = 5000
• MAX_LOG_BATCH_ENTRIES = 1500
• MAX_LOG_BATCH_LENGTH = 150000
• SUPERVISOR_KILL_TIMEOUT_MS = 500
• SUPERVISOR_LOG_TIMEOUT_MS

• /load - Not used after node.
• /check - Heartbeat, must return 200 OK
• /execute - POST - For execution of all non-http functions
• /execute/* - For HTTP will accept arbitrary paths.

• X-Google-Status: one of crash, load_error, error to indicate why the function died.

Accept logs and kill command from worker

• _ah/kill - Notify of our need to die.
• _ah/log

  {
    "Entries": [
      {
        "TextPayLoad": "...",
        "Severity": "INFO",
        "Time": "2018-01-01T00:00:00.000Z"
      }
    ]
  }

Events from firebase don't wrap the context in a sepate field.

projects/<project name>

• providers/firebase.auth/eventTypes/user.create
• providers/firebase.auth/eventTypes/user.delete

projects/<project name>/databases/(default)/documents/<collection path>/{collectionId}

• providers/cloud.firestore/eventTypes/document.create
• providers/cloud.firestore/eventTypes/document.update
• providers/cloud.firestore/eventTypes/document.delete
• providers/cloud.firestore/eventTypes/document.write

projects/_/instances/<instance|project name>/refs/<path>/{key}

• providers/google.firebase.database/eventTypes/ref.create
• providers/google.firebase.database/eventTypes/ref.write
• providers/google.firebase.database/eventTypes/ref.update
• providers/google.firebase.database/eventTypes/ref.delete

projects/<project name>/topics/<topic name>

• google.pubsub.topic.publish

projects/<project name>/buckets/<bucket name>

• google.storage.object.finalize
• google.storage.object.metadata_update
• google.storage.object.delete
• google.storage.object.archive