It doesn't matter what it is or what is in it, it just needs to exist
on disk so that platform code can touch it.

Does so by running the edxapp role just to generate config (as
controlled by check-migrations.sh's use of the --tag) and then
the tasks in the playbook are hardcoded to always run, regardless
of the tag.  Because the edxapp play has run, the tasks have
access to the various config variables containing DATABASES
and can tell what DBs are in use.

Skip the read replica

The unique constraint was removed in our fork, this sed is a no-op
https://github.com/edx/django-openid-auth/commit/ea49a01d85cb2fccc5fe34f3d0e042dc1deaf8e8

https://docs.djangoproject.com/en/1.8/ref/django-admin/#syncdb

Correct typo.  Spelling is hard.

Add the EDXAPP_CFG_DIR so that yaml templates get created in the correct location.

`repo.txt` and `local.txt` don't always exist.

Add set -e so the script fails if any command in the script fails.

check for var definitions

create configs for migration

fixing ansible command

setting extravars

run migrations using a new ansible play

missing spaces

removing venv var

adding verbose flag to ansible run for migrations

fixing newline output

add db_dry_run var

add db_dry_run logic

facts not needed for migration

fixing name