Short Notes on AWS

Can't connect to EC2 instance

The obvious 2 problems with incoming requests, that are outside of AWS's scope:

• check the instance's firewall
• check that the app is listening to all incoming (0.0.0.0/0 or your IP, not just 127.0.0.1)

On the AWS side, check the following:

• make sure the Elastic IP is associated with the instance
  • find the instance in the EC2 > Instances
  • look under Description tab, Elastic IP
  • if it's not, go to EC2 > Elastic IPs
  • choose Elastic IP from the list (or, allocate new on) that is not associated with any instance
  • choose Actions > Associate address, and associate it with the instance
• make sure Security group permissions allow the connection
  • go to EC2 > Security Groups
  • select the security group (you can find which security group instance is in in the list on EC2 > Instances page, last column)
  • on the Inbound tab, check that your protocol is enabled for Source 0.0.0.0/0 (or from your IP)
• make sure your Internet Gateway is connected to your VPC
  • make sure the Internet Gateway is attached to your VPC, under VPC > Internet Gateways > Summary tab
  • go to VPC > Route Tables, select route table for your VPC
  • under Routes tab, make sure that route with destination 0.0.0.0/0, with Target being your internet gateway, exists and is Active

Authorization header being removed by ElasticBeanstalk

By default, AWS ElasticBeanstalk's WSGI server strips Authorization header from requests.

To get these back, just plug your wsgi config file through .ebextensions, adding a wsgi.authorization.config file, with the following content:

files:
    "/etc/httpd/conf.d/wsgiauth.conf":
        mode: "000644"
        owner: root
        group: root
        content: |
            WSGIPassAuthorization On

IAM Notes

• you need policy AmazonRDSReadOnlyAccess for your IAM to be able to list RDS instances

AWS Lambda (Py) Notes

For handler(event, context) function, parameters are in:

• GET parameters in event["multiValueQueryStringParameters"]
  • note: parameters are stored in arrays, the lambda's parser correctly presumes that there may be multiple values; e.g. event["multiValueQueryStringParameters"] = {"param": ["value"]}
• path parameters in event["pathParameters"]
  • note: path parameters are specified in the SAM yaml, under Path property, as Path: /v1/user/{user_id}/data/
• POST/PUT body in event["body"]
  • note: stored as string, you have to json.loads() or similar

boto3 Snippets

Generate presigned S3 URL

s3 = boto3.client('s3', config=botocore.client.Config(signature_version='s3v4', region_name=BUCKET_REGION))
resp = s3.generate_presigned_url('get_object', Params={'Bucket': BUCKET, 'Key': KEY}, ExpiresIn=SECONDS)

AWS Chalice

Prepare your virtualenv, and install boto3 and Chalice in it.

$ virtualenv ~/.virtualenvs/test-venv
$ source ~/.virtualenvs/test-venv/bin/activate
(test-venv) $ pip install boto3 chalice

Connection to RDS

Lambda's are typically not restricted to VPC, while RDS is strictly tied to it. You need to assign your Chalice Lambda's to the same subnets as the RDS, and the same security group. Chalice does almost all of this for you!

You just need to copy all subnet ID's used by your RDS, and the security group ID.

# .chalice/config.json:

{
  "version": "2.0",
  "app_name": "chalice-test",
  "environment_variables": {
    "var": "value"
  },
  "layers": ["arn:aws:lambda:..."],
  "stages": {
    "dev": {
      "api_gateway_stage": "api",
      "subnet_ids": [
        "subnet-...", ...
      ],
      "security_group_ids": [
        "sg-..."
      ]
    }
  }
}

Send and Download Binary Files

There are two sides of this story, influencing both your Chalice code, and influencing the receiving code (your frontend app).

On the Chalice side, you simply use the Response object with some of the registered binary types. I typically just use application/octet-stream...

@app.route('/binary-data')
def bin_echo():
    raw_request_body = app.current_request.raw_body
    return Response(body=binary_data,
                    status_code=200,
                    headers={'Content-Type': 'application/octet-stream'})

For more info see the Chalice documentation

On the frontend side, you need to make sure API Gateway does not mess you up - you need to specify Accept header of the binary type (not necessarily of the same type as the one returned from Chalice). If you specify */* as the accepted type, you'll receive base64 encoded data.

Building a Python Layer

We'll be using docker Amazon Linux image.

Below, commands starting with $ are run on your machine, while those starting with bash-4.2# are run within the docker container.

$ cat docker-compose.yaml 
version: "3"
services:
  amzlinux:
    image: "amazonlinux"
    command: "sleep infinity"
    volumes:
      - ./:/host

$ docker-compose up
Creating network "layer-example_default" with the default driver
Pulling amzlinux (amazonlinux:)...
[...]
Attaching to layer-example_amzlinux_1

$ docker ps
CONTAINER ID        IMAGE               COMMAND             CREATED             STATUS              PORTS               NAMES
a5107c00ed35        amazonlinux         "sleep infinity"    10 seconds ago      Up 8 seconds                            '''layer-example_amzlinux_1'''

$ docker exec -it layer-example_amzlinux_1 /bin/bash

bash-4.2# yum -y update
>>> bash-4.2# yum -y groupinstall "Development Tools"
>>> bash-4.2# yum -y install Cython

bash-4.2# yum -y install python3-pip.noarch zip
[...]
Installed:
  python3-pip.noarch 0:9.0.3-1.amzn2.0.1

bash-4.2# cd /host/

bash-4.2# mkdir -p sqlalchemy-layer/python

bash-4.2# pip3 install sqlalchemy -t sqlalchemy-layer/python/
[...]
Successfully installed sqlalchemy-1.3.11

bash-4.2# pip3 install psycopg2-binary -t sqlalchemy-layer/python/
[...]
Successfully installed psycopg2-binary-2.8.4

bash-4.2# cd sqlalchemy-layer/
bash-4.2# zip -r aws-sqlalchemy-layer.zip python/ -x \*.pyc

Now you can upload the aws-sqlalchemy-layer.zip as a Layer through AWS console.