[Devel] Making Bugs be able to check up on themselves

[Asheesh Laroia]

As Jack and I were sketching out the design of the asynchronous bug 
importer at http://piratepad.net/async-bug-importer , we decided that Bug 
objects should store enough information so that they can refresh their 
data through a method call.

This is the One True Way to fix the Nagios warnings and the fact that we 
sometimes show out-of-date bug information -- it will be super easy to 
loop over the Bug objects and ask them to refresh themselves.

Jack filed a bug about this: https://openhatch.org/bugs/issue368

Here are my current thoughts about it:

* There are only a few bug trackers that create really stale data. 
Two, actually: Our Mozilla Bugzilla importer, and the OpenSolaris 
bug tracker.

* I'm going to write this as a patch series and ask for review from 
someone else. And I'm not going to deploy until I get someone else's 
sign-off.

* Once we deploy it, I think we should adjust the Nagios warnings to add 1 
day -- so it gives a WARNING if there are bugs older than 2 days, and a 
CRITICAL notification if there are any older than 3 days.

Once we land Jack's burgeoning work on doing the bug tracker imports 
more efficiently (anynchronously), then we should bump that back down to 1 
day for WARNING and 2 days for CRITICAL.

DATA

Here is how I checked up on which bug trackers are leaving stale data 
in the database. I did './bin/production shell_plus' on the deployment, 
and wrote the following code interactively:

In [1]: import datetime
In [2]: two_days_ago = (datetime.datetime.now() -
    ...:                     datetime.timedelta(days=2))
    ...:
In [4]: very_stale = Bug.all_bugs.filter(
    ...:         last_polled__lt=two_days_ago)
In [5]: project_names = sorted(set([b.project.name for b in 
very_stale]))
In [6]: project_names
Out[6]: [u'Firefox', u'Mozilla Core', u'OpenSolaris OS/Net', 
u'mozilla.org']

-- Asheesh.