An experiment to test RDS's behaviour during upgrades
1. Start an RDS instance
2. Start an RDS replica
3. Start pinging INSERTs into the database, once per second
4. Follow them with a SELECT on the master and replica
5. Click the upgrade button
6. Analyse *everything*
Questions: How does this change if a multi-AZ master is used?
## Setting up
$ cd terraform
$ export aws credentials
$ terraform apply
## Doing the test
$ pipenv shell
$ export DATABASE_URL = ... <-- From the terraform output, use the master
$ ./manage.py migrate
$ ./manage.py write_to_db
In another shell
$ export DATABASE_URL = ... <-- From the terraform output, master or replica
$ ./manage.py read_from_db
When ready, make a change to the database, e.g. change the instance class