[02:30] <bigjools> jtv: I realised why our tests are so slow to start up - it's doing migrations.  I wonder if we can avoid them for test runs?
[02:30] <jtv> We can, but it's one of the two things that are hard.
[02:30] <jtv> We can "cache the schema."
[02:32] <jtv> I'm not sure but the South test runner may be doing this for us already, between tests in a single run.
[02:32] <jtv> Two ways of doing it spring to mind: template database, or SQL dump.
[02:33] <jtv> When you create a database in postgres, it basically copies a template.  You can choose the template, but of course it defaults to the built-in template.
[02:37] <bigjools> we could create a template every time a schema change is checked in
[02:37] <bigjools> create/modify whatever
[02:38] <bigjools> jtv: did you see my email to maas-devel about migrations?
[02:38] <bigjools> backporting thereof
[02:38]  * jtv looks
[02:39] <bigjools> I'm never quite sure how South is going to barf at me
[02:39] <bigjools> I can add a new migration in 1.6 as allenap suggested, but I think it would need to go in trunk as well, which negates its usefulness
[05:11] <bigjools> jtv: any ideas with this? https://bugs.launchpad.net/maas/+bug/1354309
[06:46] <jtv> bigjools: first impression is it's an excellent bug report.  Looking into it.
[06:58] <jtv> Infuriating that the traceback barely contains a hint of our code.  Hard to figure out what's going on.
[08:23] <allenap> bigjools: I forgot all about the fact that people running 1.5 will want to upgrade at some point, therefore we need continuity in migrations.
[08:24] <bigjools> allenap: pesky upgraders
[08:24] <bigjools> allenap: while you're here
[08:24] <bigjools> periodic lease uploading service - I think I need a new RPC for that since we want to replace the API call for this, right?
[08:25] <bigjools> at least I don't think we can shortcut this because pserv doesn't have the cluster secrets
[08:26] <allenap> bigjools: Once we have mutual authn between clusters and regions (using TLS) we can have a RPC call to just ask for the secrets.
[08:26] <allenap> That might be quicker right now.
[08:26] <allenap> Seeing as we need the authn bits.
[08:26] <bigjools> allenap: mmyes.  arse
[08:26] <bigjools> ok, how do we do this... let's chat on the hangout
[08:26] <allenap> Okeydoke.
[15:41] <rvba> gmb: I see metadataserver/__init__.py still uses Python's logging module directly instead of using the utility you've added during the sprint… is that okay or should we change that?
[15:43] <rvba> gmb: the CI is failing with a different error now: we expect the images for i386 to be imported and they aren't anymore (I'm guessing someone changed the default set of images to be imported).
[15:45] <rvba> blake_r: ping
[15:48] <allenap> rvba: Do you know why I’d get byte strings from MACAddress.mac_address instead of a MAC instance?
[15:55] <allenap> rvba: Don’t worry, I’ll just do unicode(instance_of_MACAddress) which seems to have been created to deal with this.
[15:56] <gmb> rvba: We should change the logging thing. I’ll do a branch now. I’ll look into the CI issue, too.
[15:56] <rvba> allenap: all right.
[15:56] <rvba> gmb: see the email I sent (to blake_r).
[15:57] <gmb> rvba: Okay, thx
[16:10] <rvba> gmb: bzr grep "import logging"
[16:11] <rvba> says there are still a couple of places that use logging directly
[16:26] <roaksoax> so shouldn't all the branches we are landing noew include proper logging?
[16:26] <roaksoax> are we doing that?