luci migration app: Ensure flaky retries of luci builds affect speed and/or correctness |
|||
Issue descriptionluci migration app: Ensure flaky retries of luci builds affect speed and/or correctness For example, see luci version of android_arm64_dbg_recipe in https://chromium-review.googlesource.com/c/chromium/src/+/905424/17 It failed, retried, and succeeded. If the first failure doesn't affect the luci analysis as reported by the migration app (https://luci-migration.appspot.com/masters/tryserver.chromium.android/builders/android_arm64_dbg_recipe/), I believe it should because the source of the flake might only apply to luci bots. (And if that's the case, we should definitely be aware of it.)
,
Feb 15 2018
FTR, the reason this is worth bringing up is that android_arm64_dbg_recipe is apparently 100% correct: https://luci-migration.appspot.com/masters/tryserver.chromium.android/builders/android_arm64_dbg_recipe/ Despite one of its bots incorrectly failing *many* builds: https://chromium-swarm.appspot.com/bot?id=swarm994-c4 Root cause is a broken deployment on that bot (I think), but that kind of systematic failure should have a way of bubbling up.
,
Feb 15 2018
,
Feb 22 2018
Maybe we should compare flake rates between buildbot and LUCI. Perhaps that's harder to collect than speed since we don't already calculate and store it, though. Maybe once we have the BQ data we can just put together some queries to spot check this.
,
Feb 22 2018
Using BQ data for checking sgtm. Is there a bug I can follow for getting the table(s) spun up and populated?
,
Feb 22 2018
big 788034 is about completed_builds for Buildbucket builds |
|||
►
Sign in to add a comment |
|||
Comment 1 by tandrii@chromium.org
, Feb 15 2018