Essentially a rewrite from the code here :

https://aws.amazon.com/articles/high-availability-for-amazon-vpc-nat-instances-an-example/

We needed big changes in order to avoid using :

• amazon-tuned centos AMIs (and the API helpers)
• chunks of messy shell scripts
• centralized and maintainable configuration

The script rely on requests and boto3. Python 3.6 was used for writing and testing

pip install -r requirements.txt
python3 monyt.py config_monyt.json

The configuration is done through the json file.

The script ping its peer, if it detects it down it will modify the route tables

The program automagically search for instances which 'Name' tag matches '-nat-' (tag and pattern are configurable in the config_monyt.json file)

On your nat-instances, you will want to have the program launched upon startup.

Either:

• Write an init script relevant to the system you are running (initV, rc, systemd, ...)
• add a line such as : screen -d -m python3 /path/to/monyt.py /path/to/config.json to /etc/rc.local (if you chose this option, the $PWD will be '/', the log will be found there by default (you can customize it in the json file))

Document the policies needed in order to avoid setting the aws-access-key being over-privileged