yellow team mailing list archive

[Brad Crittenden <brad.crittenden@xxxxxxxxxxxxx>]

Under LXC I found the log at /var/log/juju/unit*.log

See dev.launchpad.net/yellow/Juju for other tips, such as some precise work arounds. 

bac



On Jan 23, 2012, at 17:02, Gary Poster <gary.poster@xxxxxxxxxxxxx> wrote:

> First, a quick note. Thanks to Serge and Andy, Precise now has our desired (and fixed) version of lxc-start-ephemeral (I verified), and the script does not cause a XWindow crash (I verified with a custom kernel on Oneiric, and it is supposed to be released on Precise now too).
> 
> Next, as a reminder, here's our high level plan.  I probably should put this on a wiki page...
> 
> Stage 1: get a repeatable deployment story for our deliverable: generic buildbot master and slave charm configurable to run LP tests in parallel.
> 
> Stage 2: Run experiments to determine how many cores to advise IS to purchase for the final buildbot slaves.
> 
> Stage 3: Run tests repeatedly to find errors and fragilities.  When tests run reliably enough, for some as-yet-to-be-determined definition of that, stop and give the project to IS to be deployed.
> 
> We're in stage 1. :-)
> 
> So, now the meat of my mail.  If you are looking for a task tomorrow, what could you do?  Thanks to Brad for helping me brainstorm these. These are in no particular order.
> 
> 1) Get precise running on metal, if you haven't already.
> 
> 2) Get the buildbot master Juju charm that Brad and Graham have been working on working on your system with lxc. lp:~yellow/launchpad/buildbot-master
> 
> 3) Make sure that our story (setuplxc, parallel ephemeral tests) works in kvm.  We are supposed to be doing step 2 in a kvm.  Francesco has been testing setuplxc in a kvm, so testing parallel tests and the ephemeral lxc script themselves is the main job here.
> 
> 4) [Check with Brad for status of this before starting] https://bugs.launchpad.net/juju/+bug/886207 is causing problems, particularly because the work around (to look on the local log) does not seem to work when you use lxc.  Maybe try to fix it?  Could be a slack task.
> 
> 5) Brad is working on making the buildbot master charm check out its configuration from a repo.  Maybe help him?
> 
> 6) We need a buildbot slave recipe.  It needs to be able to accept the setuplxc script as an argument somehow and know how to run it to initialize the slave (see also the next idea).
> 
> 7) Brad and I agreed on this pattern for build slaves: build masters accept a master.cfg that is templatized for adding in build slave names and build steps.  When juju adds a buildbot slave, the master generates a password to use; the slave tells the master what steps it wants to run and a name suggestion; the master uses the name or a variant if it is already in use; the master substitutes the slave's name, password and steps into the templatized config; the master reloads config; the master informs the slave of the name and password, and the slave sets up the configuration for itself and starts up, to start a conversation with the master.  You can work on setting up this relationship part of the juju charm (or work on your alternate proposal for how things should work :-) ).  You could start on the buildbot master part of this handshake while waiting on someone else to build the basic buildbot slave charm, maybe?
> 
> 8) Pure slack time task: help get Launchpad's tests passing on Python 2.7, in preparation for developing and deploying on Precise.
> 
> Have fun, make a kanban card for what you do, and please collaborate! Timezones will naturally separate the Europeans and Americans, and we must fight that sometimes, but don't fight convenience all the time. :-) Maybe Graham and Francesco should collaborate on your next task, and a couple of Americans ought to collaborate.
> 
> Gary
> 
> -- 
> Mailing list: https://launchpad.net/~yellow
> Post to     : yellow@xxxxxxxxxxxxxxxxxxx
> Unsubscribe : https://launchpad.net/~yellow
> More help   : https://help.launchpad.net/ListHelp