telemetry: run_benchmark should enable "stay awake" and disable rotation |
|||||||
Issue descriptionToday I had some issues running a benchmark locally. Tuns out that tools/pers/run_benchmark is not setting the "stay awake" setting, nor the force portrait mode. I guess they happen in the bot scripts, but would be useful to have them been done in run_benchmark as they improve reliability. P.S: couldn't figure out if bugs for run_benchmark should be filed here or in catapult. run_benchmark seems to be still living in chrome.
,
Apr 13 2016
,
Apr 13 2016
Yeah I am pretty sure we do on the bots. Just think this should happen in general, or people will fail to reproduce benchmarks locally.
,
Apr 13 2016
+dtu, aiolos, rnephew We have a Q2 goal to normalize a lot of the setup in the various recipes that run telemetry. But we didn't think about this part of running locally. Does anyone know why this is in the recipe but not locally? (And yes, crbug is the right place for run_benchmark bugs)
,
Apr 13 2016
+nednguyen We don't yet have a good way to differentiate what setup steps happen in Telemetry vs outside of Telemetry. In general, I don't think there's a way to guarantee a local run can reproduce the environment of a bot run (other than what Ned called a "best effort"), so I'm not yet convinced we can even support local runs.
,
Apr 13 2016
,
Apr 13 2016
I'm not 100% sure if it will be identical, but if you run the provision_devices script in build/android you can get pretty close to on bot setup.
,
Apr 13 2016
Yeah, from our discussion last time, for environment preparation step do not have severe side effects, we should move it to telemetry: enable stay awake & disable rotation seems ok in this case. For the ones that have severe side effect like killing all the background services/wait few hours for the device to cool off, we shouldn't put them in telemetry.
,
Apr 14 2016
provision_devices is a pretty big hammer (iirc it wipes the device by default). I tend to agree with #8, disabling rotation and forcing awakeness seems to belong more to the telemetry / run_benchmark world than to the bot provisioning.
,
Apr 14 2016
Agreed with comment 8.
,
Apr 14 2016
We don't have anyone on our team who can take this at the moment. So let's put this in our telemetry backlog for now.
,
Apr 14 2016
,
Apr 15 2016
I think this would actually be part of our Q2 OKR to refactor the recipes so they are more consistent (moving things from recipe->telemetry when it makes sense sounds like part of that).
,
Sep 13 2017
|
|||||||
►
Sign in to add a comment |
|||||||
Comment 1 by perezju@chromium.org
, Apr 13 2016