Deployment

In order to run applications independently of the computing environment, containerization is the most common solution. There are many ways to deploy containerized applications. Here we provide examples and tips to help you getting started.

Docker Compose

Docker Compose is a very popular tool for defining and running multi-container Docker applications. With Compose, you use a YAML file to configure your application’s services.

Using Compose is basically a three-step process:

• Define your app’s environment with a Dockerfile so it can be reproduced anywhere.

• Define the services that make up your app in docker-compose.yml so they can be run together in an isolated environment.

• Run docker compose up and the Docker compose command starts and runs your entire app.

Deployment with Docker Compose

A basic Docker Compose template for orchestrating a Flask application & a Celery queue with Redis can be found at amplpyfinance/deployment/docker.

Jobs submitted to the flask application are sent to the queue and then processed by one of the optimization workers (docker containers with AMPL and solvers installed) where amplpyfinance is used to solve optimization problems.

Project structure

• docker-compose.yml: definition of the services involved in this Docker application.

• api/

  • Dockerfile: Dockerfile for the Flask application.

  • app.py: Flask application.

  • requirements.txt: Python requirements for the Flask application.

• queue/

  • Dockerfile: Dockerfile for Celery and its workers.

    • This Dockerfile includes the steps to install AMPL and solvers since this needs to be available in the workers.

  • tasks.py: definition of Celery tasks and startup handler (used to activate the AMPL license when a new worker starts).

  • requirements.txt: Python requirements for the Celery queue and workers.

• client.py: Python client to submit requests to the application.

Installation

git clone https://github.com/ampl/amplpyfinance

AMPL and solver licenses

You can use the demo version that is subject to size limits or any cloud license (including AMPL CE licenses). You can specify the license by setting the environment variable AMPLKEY_UUID.

export AMPLKEY_UUID=xxxxxxxx-xxxx-Mxxx-Nxxx-xxxxxxxxxxxx

You can also create a .env file in this directory with environment variables such as AMPLKEY_UUID.

AMPLKEY_UUID=xxxxxxxx-xxxx-Mxxx-Nxxx-xxxxxxxxxxxx
COMPOSE_PROJECT_NAME=amplpyfinance
...

Build & Launch

$ cd amplpyfinance/deployment/docker
$ export COMPOSE_PROJECT_NAME=amplpyfinance
$ docker compose up -d --build

Enable hot code reload

In addition to docker-compose.yml, we provide dev.yml to override some specifications in docker-compose.yml to address development needs such as hot code reload.

$ export COMPOSE_PROJECT_NAME=amplpyfinance
$ docker compose -f docker-compose.yml -f dev.yml up --build

This will expose the Flask application’s endpoints on port 8080 as well as a Flower server for monitoring workers on port 5555

To add more workers:

$ export COMPOSE_PROJECT_NAME=amplpyfinance
$ docker compose up -d --scale worker=5 --no-recreate

To shut down:

$ docker compose down

To change the endpoints, update the code in api/app.py

Task changes should happen in queue/tasks.py

Submit jobs to the containerized application

After running docker compose up check the endpoints with docker compose ps to make sure they are running and to see the ports being used.

You can then use client.py with python client.py 127.0.0.1:80 to submit jobs. When running locally the IP address is 127.0.0.1 and the port should be 80 or 8080 (if dev.yml is being used).

You can check the status of the jobs by accessing the Flower server that uses port 5555. You can check in your browser http://127.0.0.1:5555/ when running locally.

Deploying on AWS using Docker Compose

Deploying Docker containers on ECS can be done with docker compose as follows:

• First you need to build and publish your docker images:

  • Build the images with docker compose build. Note: you will need to rename the images in docker-compose.yml in order to upload to your own account.

  • Upload the docker images docker compose push.

• You can then deploy the application as follows:

  • Create a docker context for ECS with docker context create ecs myecscontext.

  • Select the context to use with docker context use myecscontext.

  • Deploy with docker compose up -d.

  • Check the endpoints with docker compose ps in order to know the the IP addresses and ports being used.

• Stop the deployed application docker compose down (if you forget this step the application will keep running and you will keep being charged for the resources being used).

You can get to the default context (i.e., local deployment) with docker context use default.

You can read more about this type of deployment at Docker Compose: From Local to Amazon ECS.

Deploying on Azure using Docker Compose

Deploying Docker containers on Azure can be done with docker compose as follows:

• First you need to build and publish your docker images:

  • Build the images with docker compose build. Note: you will need to rename the images in docker-compose.yml in order to upload to your own account.

  • Upload the docker images docker compose push.

• You can then deploy the application as follows:

  • Create a docker context for ECS with docker context create aci myacicontext.

  • Select the context to use with docker context use myacicontext.

  • Deploy with docker compose up -d.

  • Check the endpoints with docker compose ps in order to know the the IP addresses and ports being used.

• Stop the deployed application docker compose down (if you forget this step the application will keep running and you will keep being charged for the resources being used).

You can get to the default context (i.e., local deployment) with docker context use default.

previous

DiscreteAllocation