Substantially similar to the certs role developed (in paralell) that
exists on master, but this one uses pre-Croissant directory structure
and variable conventions. Modeled on the croissant-era notifier role.
Useful as a bridge to midterms, when Stanford can afford to port to the
new ways of doing things.

Includes updates to logging-related tasks for ansible and datadog. These
are ongoing problems for us at the moment.

add host for mysql

PHP / PHPMYADMIN role

Add logrotation to Stanford datadog setup

* Datadog logs now get rotated like other system logs
* nginx logs get rotated to have group syslog, so the datadog agent will
  continue to be able to read them

* Moving test() contents into doctests
* Adding comments to the methods explaining what they're used for

Stanford-oriented Datadog updates

* Datadog explicitly sets hostname in style Stanford uses (relies on
  /etc/hostname being accurate, requires reboot)
* Enables nginx status check module integration for datadog
* Installs nginx access log processor that counts HTTP response codes
* Updates to lots of other places to make sure that their logs appear
  where we expect
* Has sample code showing how to emit events from log contents which make it
  into the Datadog event stream
* Separation of stage and prod datadog configurations

install edxapp code sandboxing

* Changed the way xqueue does basic auth configuration so that it would
  work with the old httplib on our prod ora instances.
* this breaks the .org release's method of doing basic auth, but that
  doesn't matter for Stanford because we do it differently

Conflicts:

	playbooks/roles/ansible_debug/tasks/main.yml

* xqueue user creation fails mysteriously sometimes. We don't actually
  care if the user creation works because most of the time, the user has
  already been created. So, ignore these errors.

The ORA machinery gets gummed up when it gets very large submissions and
throws a lot of 500s.

Co-Authored-By: Jason Bau <jbau@stanford.edu>

Conflicts:
	playbooks/roles/edxapp/defaults/main.yml

logrotate is not a service

openid workaround needs to happen before the syncdb/migrate

Fixes to the xqueue and xserver roles.

Create "pinned" files to allow upgrading plugins bundled with Jenkins.

Upgrade git version used on the Jenkins worker

Fix Jenkins version to last known stable.

Fixed bad YAML and locked db to specific variable configured version