benchmark_duration alerts include one that is less than 5 minutes regression |
|||
Issue descriptionWe planned to only alert on benchmark duration regression that are more than 5 minutes. Simon: can you help triaging this?
,
May 8 2017
It kinda looks like benchmark_duration is reported in seconds not minutes, so the anomaly config
{
"min_absolute_change": 5.0
}
Would alert on changes of 5s or more. I can switch that to 300, just want to confirm first that the benchmark_duration is in seconds.
,
May 8 2017
benchmark_duration is in minutes, I am pretty sure: https://github.com/catapult-project/catapult/blob/master/telemetry/telemetry/internal/story_runner.py#L375 Though what may happen here is the dashboar automatically convert minute to seconds?
,
May 8 2017
Ah my bad, yes it does look like it's in minutes, I'll dig in a bit and see what's up then.
,
May 8 2017
One thing I noticed digging through the datastore is that, the example link has 2 benchmarks, both speedometer, which alert on < 5 mins. The anomaly config on TestMetadata('ChromiumPerf/win-high-dpi/speedometer/benchmark_duration') is actually assigned to the v8-speedometer anomaly config. The code in TestMetadata that assigns anomaly configs grabs all configs and sorts alphabetically, but is case sensitive so 'benchmark_duration' would fall after 'V8-speedometer' due to case.
,
May 8 2017
Perhaps we should prefer the more specific pattern matching over the general one as a potential fix.
,
May 11 2017
The following revision refers to this bug: https://chromium.googlesource.com/chromium/src.git/+/d997518a95d124b2deebb67e37c96cff7fe5c1c2 commit d997518a95d124b2deebb67e37c96cff7fe5c1c2 Author: catapult-deps-roller@chromium.org <catapult-deps-roller@chromium.org> Date: Thu May 11 17:11:56 2017 Roll src/third_party/catapult/ 23c9f59f4..59e58189c (1 commit) https://chromium.googlesource.com/external/github.com/catapult-project/catapult.git/+log/23c9f59f4c6b..59e58189c55d $ git log 23c9f59f4..59e58189c --date=short --no-merges --format='%ad %ae %s' 2017-05-11 simonhatch Dashboard - Choose anomaly config which most specifically matches. Created with: roll-dep src/third_party/catapult BUG= 718981 Documentation for the AutoRoller is here: https://skia.googlesource.com/buildbot/+/master/autoroll/README.md If the roll is causing failures, see: http://www.chromium.org/developers/tree-sheriffs/sheriff-details-chromium#TOC-Failures-due-to-DEPS-rolls CQ_INCLUDE_TRYBOTS=master.tryserver.chromium.android:android_optional_gpu_tests_rel TBR=sullivan@chromium.org Change-Id: Icf38804f7c158bfc919f3e72f124aa576f8f53f6 Reviewed-on: https://chromium-review.googlesource.com/502633 Reviewed-by: <catapult-deps-roller@chromium.org> Commit-Queue: <catapult-deps-roller@chromium.org> Cr-Commit-Position: refs/heads/master@{#470985} [modify] https://crrev.com/d997518a95d124b2deebb67e37c96cff7fe5c1c2/DEPS
,
May 11 2017
Next step would be to re-put() the affected TestMetadata's.
,
May 11 2017
Ok they've been resaved and I checked a few in datastore and their configs have switched over to benchmark_duration. Can keep an eye out for the alerts and see if anything is alerting on < 5 mins in the future.
,
May 11 2017
Thanks Simon!
,
Jul 18 2017
,
May 18 2018
I think I just forgot to close this, closing now. |
|||
►
Sign in to add a comment |
|||
Comment 1 by nedngu...@google.com
, May 5 2017