Simple Docker Containerization rekogGetter for WWE IG Username rekognition results

Etienne Jacquot - 09/16/2020

Getting Started

• This build's dockerfile contains input value which is the WWE IG Username... Following previous I continued by using romanreigns for this containerization testing!

• Applying principles of commentGetter containerization, for each URL (final url on https://imghost.asc.upenn.edu) we can efficiently get image binary & then pass this value to AWS label rekog w/ all attributes

---

Confirm your AWS credentials

• You must configure your default AWS credentials in order to run AWS Rekognition API calls!

  • Contact your AWS sysadmin for your access & secret tokens at [email protected]

• Once you have your AWS creds, please make sure to save them to ./configs/.aws/config with the following formatting

  • More information on various AWS config file here

# Confirm your credentials are correct!
cat ./configs/.aws/config

[default]
aws_access_key_id=AKIA...
aws_secret_access_key=P9Zd...96+dq

# If not, you must edit these
vim ./configs/.aws/config

• You'll notice below that this file path (absolute inside of the rekogGetter container) is included as an --env flag for the docker run command!
  • This is singificant because we must avoid hardcoding creds in the GetRekognition script (this is always a bad practice!). We also must ensure that the ../gitignore for ther repo excludes this config file so it is not getting pushed!

# Confirm your config file is added to the gitignore!

echo "docker_rekognition/configs/.aws/config" >> ../.gitignore

cat ../.gitignore

---

Build container for WWE IG Username

• This only runs for label (more info here: https://boto3.amazonaws.com/v1/documentation/api/latest/reference/services/rekognition.html#Rekognition.Client.detect_labels)

• Dockerfile containers input of instagram username ... Should this be one URL for one container, or does one Username (many URLs as rows lambda x:) to one container)

# Check & confirm dockerfile input value for WWE IG Username
cat dockerfile | grep python3

docker build -t wwe_instagram_rekog -f ./dockerfile .

---

Run container for WWE IG Username

• Make sure to set your docker mount, I am running this on my macbook w/ docker...

• Docker flags for run command:

  • -it for docker shell to display results, helpful for testing (could use -d for detached)
  • --env for the AWS_CONFIG_FILE
  • -v for the Docker volume mount on your localhost!
  • --network=host for Docker on MacOS, this is known issue
  • --name to set the container name as rekogGetter

# Specifying environment variable for AWS configs so they are not hard coded

docker run -it --env AWS_SHARED_CREDENTIALS_FILE=/app/configs/.aws/credentials/ -v ~/Documents/Bitbucket/wwe-instagram-rekogcomments/docker_rekognition/data/:/data/ --network=host --name=rekogGetter wwe_instagram_rekog:latest

---

NOTES & CHANGES FOR NEXT TIME:

• Based on the previous notebooks there is a lot of room for improvement for my python Rekogntion for WWE comments...

• Passing image_binary for label_response allows you to create ['AWS_Label_Rekognition'] column, but can you return multiple values for lambda x: to create multiple columns? Just thinking of combining all rekognition API calls for one image binary!

1. one username to one container OR one URL to one container?
2. incorporating rekog detection for more than just label
3. AWS creds are in very poor fashion coded in py script...
   1. learn one of the options described by AWS here --> https://boto3.amazonaws.com/v1/documentation/api/latest/guide/credentials.html
   2. env variables in dockerfile? config in .aws/config somewhere from ./configs/config ...? https://vsupalov.com/docker-arg-env-variable-guide/ for dockerfiles you use ARG and ENV to make the environment variable available to the final container!
4. Ultimately, scalability for WWE IG users w/ many posts

• This docker build demonstrates a relatively simple process