On Tue, 10 Jun 2014 08:51:29 +0000 "Chisanovici, IonutX" ionutx.chisanovici@intel.com wrote:
Actually, the setup is like this (I have detailed this also):
- One of Fedora/Ubuntu/OpenSuse/CentOS machine which should be as
device in lava server (on this machine will run the tests). I have to have these os'es and not to have a lava prebuilt image.
The OS can be inside a VM (and therefore deployed inside LAVA) and it only needs a LAVA overlay to run the tests. The image would be created by you, to your needs, just downloaded by LAVA. It wouldn't have to be an image built by Linaro. This would be a KVM device using a purpose-built image. (Using KVM means you only need one machine, as long as that machine has enough cores and enough RAM.) Alternatively, LAVA can just use a connection to one existing machine for each OS, via the dummy devices.
- The above
machine has to send the test results to lava server some how (JSON/YAML) 3. See the test results in lava dashboard.
Without the rest of the build log, the results aren't going to be useful.
I'm pretty sure that this scenario can happen and can be implemented using lava server.
Probably via the dummy-ssh device. The machine would already be running, a test would login, run the build so that the logs appear in LAVA and then collect the results as normal. No need to have external test results, just openssh-server running.
Any command used to generate the results externally can be executed within LAVA - all it needs is a connection to that box. If you don't want a cleanly booted image, then use a connection to a dirty environment which is already running but a clean environment is preferable.
LAVA can support multiple nodes per test job, so you could do all the tests at the same time.