Many power metrics that aren't story:power_avg have stopped reporting recently |
||
Issue descriptionHere's an example trace that was generated recently: https://console.developers.google.com/m/cloudstorage/b/chrome-telemetry-output/o/trace-file-id_46-2017-03-30_14-18-43-86130.html Opening up the side panel, there's only a story:power_avg metric for the powerMetric, even though there's a clear load, indicating that there should be at least a load:energy_sum metric and and after_load:power_avg metric.
,
Mar 31 2017
Strange. There doesn't seem to be anything suspicious in the revision range of the stoppage: https://chromium.googlesource.com/chromium/src/+log/f05540d87c83d707b4e9d7a00f29ebd423b9cfb8%5E..1385715e4bbb9aa74264578e5dba49890c49c12d?pretty=fuller
,
Mar 31 2017
It looks like the problem is that the power trace isn't covering the entire Chrome bounds, and our response to this has always been to not report more detailed power metrics if this is the case. I wonder why all of a sudden it stopped covering the entire Chrome bounds?
,
Mar 31 2017
Somewhere between in the range of 460358-460399, this stopped working. I'm going to launch a bisect: even if it ends up that we want to change the metric somehow, I still think it's worth understanding how it broke. We haven't seen anything like this before.
,
Mar 31 2017
Started bisect job https://chromeperf.appspot.com/buildbucket_job_status/8983596314422885152
,
Mar 31 2017
=== BISECT JOB RESULTS === Bisect failed for unknown reasons Please contact the team (see below) and report the error. Bisect Details Configuration: android_nexus5X_perf_bisect Benchmark : system_health.common_mobile Metric : after_load:power_avg/load_news/load_news_wikipedia To Run This Test src/tools/perf/run_benchmark -v --browser=android-chromium --output-format=chartjson --upload-results --pageset-repeat=1 --also-run-disabled-tests --story-filter=load.news.wikipedia system_health.common_mobile Debug Info https://chromeperf.appspot.com/buildbucket_job_status/8983596314422885152 Is this bisect wrong? https://chromeperf.appspot.com/bad_bisect?try_job_id=5288391580581888 | O O | Visit http://www.chromium.org/developers/speed-infra/perf-bug-faq | X | for more information addressing perf regression bugs. For feedback, | / \ | file a bug with component Speed>Bisection. Thank you!
,
Mar 31 2017
Looking at the trace in the original comment, the power trace covers the entirety of _SystemHealthSharedState.RunStory which means that it is not stopping early. It appears that the chrome trace is running longer than it use to, not that the power trace is running for less time than it use to.
,
Apr 11 2017
Started bisect job https://chromeperf.appspot.com/buildbucket_job_status/8982624347095701744
,
May 4 2017
It looks like this started working again, so I'm going to close this as WontFix. |
||
►
Sign in to add a comment |
||
Comment 1 by charliea@chromium.org
, Mar 31 2017