1. a twistd deamon polling hg for new pushes and inserting them
into the DB; a buildbot running compare-locales, producing build
logs, which are stored as text files (and in the DB via the mbdb
app?),
2. elmo dev instance,
3. elmo stage (used as prod).
With the new boxes setup, we now have an NFS share that will be mounted
on all of the boxes. How should we go about using that share?
Should there be files produced by #1 above on the NFS that will be
shared by elmo dev, stage and prod?
- easier to maintain,
- regular DB imports from prod to dev are good enough,
- less load on hg.mozilla.org.
Or, should there be independently running instances of #1, one for dev,
and one for stage+prod?
- dev tests the infrastructure holistically,
- we could test changes to #1 (twistd, buildbot, mbdb) on dev before
we deploy them on stage/prod.
Thoughts?
-stas
--
Staś Małolepszy
mozilla.org
One instance. The bots are pretty care-intensive, and cause a good deal
of load on the upstream infra like the hg server, so we shouldn't
duplicate those.
Axel
With just one instance, how can we:
- test the buildbot,
- deploy changes to the buildbot?
If the concern is the load on hg.mozilla.org, maybe we could have the
dev instance of the buildbot poll for pushes in a special, test repo in
hg?
You'd have to make a case what that dev instance is testing how.
There are tests for the actual codepaths etc, so?
Axel