IRIS App

Sample Flask web App using Scikit-Learn to predict IRIS Flower.

Deploy Steps for AWS Ubuntu 14.04 LTS EC2 Instance

Login to AWS Instance:

ssh -i <your AWS Pem key file> ubuntu@<aws ip>

Install Python / Git

sudo apt-get update
sudo apt-get upgrade

# Install GIT
sudo apt-get install git

# Install Anaconda (Miniconda)
wget https://repo.continuum.io/miniconda/Miniconda3-latest-Linux-x86_64.sh

bash Miniconda3-latest-Linux-x86_64.sh

# To update the path (Make sure you said Yes when it asked to update path in the Miniconda install steps)
source ~/.bashrc

Download App Source Code:

# Create a Projects directory where we will keep the App source files
mkdir projects
cd projects


# Clone the Repo
git clone <your app repo>

Create Conda Environment

cd <your app directory>
# do `ls` to confirm that you see environment.yml and other files

# conda create env: 
conda env create -f environment.yml

# Activate environent
source activate iris-app-env

Run App

# Try running them manually to see if that works.  More than likely fail
python run.py

# We need to run the model script
make clean
make build_model_iris

# Now run the app
python run.py

Install Supervisor

# Install Supervisor to run our App process
apt-get install supervisor

# Create Supervisor Config
sudo vi /etc/supervisor/conf.d/iris-app.conf

# Go to insert mode by pressing `i` then Add the following config

[program:iris-app]
autostart = true
autorestart = true
command = /home/ubuntu/miniconda3/envs/iris-app-env/bin/gunicorn run:application -b localhost:8000
directory = /home/ubuntu/projects/iris-predictor-flask-app
environment = PYTHONPATH="/home/ubuntu/miniconda3/envs/iris-app-env/bin/"
numprocs = 1
startsecs = 10
stderr_logfile = /var/log/supervisor/iris-app.log
stdout_logfile = /var/log/supervisor/iris-app.log

Replace the Path with your path setup in the command. If you installed miniconda2, you may need to change miniconda3 to 2.
Replace the directory to where your application is installed.
Replace environment:PYTHONPATH to the conda environment.

Go to vi command mode by pressing esckey and:wq` to write and quit.

Run Supervisor

# Reload supervisor config
sudo supervisorctl reload

# Start all supervisor process
sudo supervisorctl restart all

# Check status
sudo supervisorctl status

If the process failed or say anything other than Running (after it says starting), then check supervisor logs in /var/log/supervisor/..

Redirect Port

Redirect Port 80 to 8000 where your app is running

sudo iptables -t nat -A PREROUTING -p tcp --dport 80 -j REDIRECT --to-port 8080

Open Browser and check if you can see your app from http://<your aws ip>.

Credits:

Template from https://github.com/sampathweb/ml-cookiecutter-starter-flask-app

wccosby/flaskML