We’ve just deployed the 2011.08 release of the LAVA components to http://validation.linaro.org. It feels like we’re getting to the point where we can start working on the really interesting features now, which is nice after what feels like months of building infrastructure. While the official release highlights are a good list of what happened, here is a summary from my point of view of the changes from what was previously deployed:
- I added a put_ex method for uploading a bundle and getting the URL of the bundle in response rather than just its SHA1.
- The main change though is a shiny new image status page.
- I changed the dispatcher to print out the URL of the result bundle (on a specific fd, so the scheduler can cleanly see it in amongst all the useful but noisy stuff the dispatcher usually outputs)
- Le Chi Thu worked on being able to define test suites to run outside the lava-dispatcher/lava-test source code.
- There were also sundry changes to make things more reliable.
- Many changes here – I spent most of the month hacking on the scheduler
- Record the link to the results as output by the dispatcher and show it on the job page.
- I made the scheduler daemon reconnect to the database if the connection fails or drops. This means we could finally change to starting the daemon on boot, without having to care if postgres was fully initialized or not.
- I changed the way the dispatcher is run to allow the scheduler daemon to be restarted while there are running jobs.
- Jobs can now be cancelled while they are in the submitted or running states.
- There is now a simple device view, and admins can ask for no more jobs to be submitted to a board.
- The main changes here were to allow the other LAVA components to provide content to put on the front page.
- The work to support out-of-tree tests had an impact on lava-test, too
- We added support for the smem, bootchart and xrestop tests, as well as the usual reliability fixes.
- The bootchart test for a while managed to overwrite the kernel for the master/recovery partition of the SD cards, which was a problem, especially when the new kernel had broken ethernet support…
What’s on for next month? Well from my side, it’s likely to be still around ‘joining the dots’, starting with enabling the android team to see the results of a test run on the build page. I think other likely developments are to be making our story for running tests on android a bit clearer, refactoring the way configuraton is done for the dispatcher (and making debian packages for it), supporting more hardware in the lab, putting the final pieces together to allow daily testing of the kernel working group output, documentation, email notifications of various things, and I’m sure various things that I’ve forgotten or haven’t cropped up yet. It’s going to be another busy month!