← Back to team overview

launchpad-dev team mailing list archive

Adopting loggerhead

 

The Launchpad team @ Canonical have recently taken a much broader view
on maintenance of their code base and dependencies. Loggerhead is one
of these dependencies, and as such we'll be more consistently
maintaining the branch of Loggerhead that we run.

While we could take a very narrow view on this and just focus on the
branch that Launchpad runs, However, after some fairly lengthy
discussions, we are going to adopt Loggerhead as a project, in much
the same way that the Launchpad team looks after the lazr-* stack of
projects - things we depend on but are used for more than just
Launchpad itself. Doing this is slightly more work for the Launchpad
team but should get more attention given to Loggerhead as a whole, by
the team responsible for the single largest deployment of Loggerhead.

Implementation wise:
 - We're going to move Loggerhead from the Bazaar project group to the
launchpad-project group on Launchpad. This is due to a Launchpad
limitation - ideally we'd just add it to the launchpad-project group
and have it participate in both. The reason this limitation matters is
that if the project is not in the
https://launchpad.net/launchpad-project/+active-reviews view (and
likewise for bugs) then the folk driving the queues to zero will not
see the Loggerhead tasks.
 - The Launchpad team will maintain the project - we'll drive bug
triage and code review in the same way we do for all the other
components we are custodians of. Michael Hudson was previously a
maintainer in his role as part of the Launchpad Code team.
 - The Launchpad team will get commit access (but subject to their
normal peer review rules)
 - As we do for other projects, we'll do releases off of trunk, and
use releases in Launchpad, rather than essentially running a fork.
 - To enable this, we're going to put the work on trunk which isn't
yet deployable in Launchpad into a new 'future' branch, and reset
trunk back to 1.18

Our short term maintenance focus is driving OOPS (both exceptions and
timeouts) in the Launchpad production environment to zero. We're
seeing 16000 a day from our Loggerhead instances, but most of them
look to be one particular problem (although that may be masking deeper
symptoms).

In the future, probably later this year (because codebrowsing is a key
component of Launchpad), we'll be doing significant work on
responsiveness and scalability. This is likely to involve reviewing
the work in the 'future' branch.

Cheers,
Rob



Follow ups