launchpad-dev team mailing list archive

[Tom Haddon <tom.haddon@xxxxxxxxxxxxx>]

On Thu, 2010-07-01 at 11:49 +1000, Martin Pool wrote:
> On 1 July 2010 01:20, Max Kanat-Alexander <mkanat@xxxxxxxxxxxx> wrote:
> >        So, I think there should be an edge codebrowse where loggerhead code
> > can be tested before it goes into production.
> >
> >        This would first require a stable branch of loggerhead that is
> > different from the trunk. I'm going to discuss that on the bzr list.
> > Once we have that branch, then production should only run the stable
> > branch.
> >
> >        Edge can be running either:
> >
> >        * The stable branch before we put a new bug fix into production
> >        or
> >        * The trunk, if we're going to be updating to trunk soon.
> 
> Sorry, I know I suggested multiple threads so I can only blame myself
> for the multiplication.

Erm, yeah, we also have an RT for this - RT#38985. Can we decide whether
we're at the stage of an RT, and if so move the discussion to there, and
if not, close the RT?

My main concern here is avoiding unnecessary complication. We already
treat the "production" codebrowse as an "edge" service in that we
rollout to it whenever there's new code. This is mainly because it needs
restarting enough anyway that it can't really be considered a "stable"
service, and if we have a "stable" and an "edge" service, we'd just be
restarting two services instead of one. I think it'd be better to
continue with the current setup until we get to the point where
codebrowse doesn't need restarting between Launchpad releases. Then we
can think about splitting it out into "production" and "edge" instances.

Thanks, Tom