New issue
Advanced search Search tips
Note: Color blocks (like or ) mean that a user may not be available. Tooltip shows the reason.

Issue 690211 link

Starred by 1 user

Issue metadata

Status: WontFix
Owner:
Last visit > 30 days ago
Closed: Aug 2017
Cc:
EstimatedDays: ----
NextAction: ----
OS: ----
Pri: 2
Type: Bug-Regression



Sign in to add a comment

1.4MB regression in system_health.memory_mobile at 447218:449231

Project Member Reported by briander...@chromium.org, Feb 8 2017

Issue description

See the link to graphs below.
 
All graphs for this bug:
  https://chromeperf.appspot.com/group_report?bug_id=690211

Original alerts at time of bug-filing:
  https://chromeperf.appspot.com/group_report?keys=agxzfmNocm9tZXBlcmZyFAsSB0Fub21hbHkYgIDg6PCppgsM


Bot(s) for this bug's original alert(s):

android-nexus5X

=== BISECT JOB RESULTS ===
NO Perf regression found

Bisect Details
  Configuration: android_nexus5X_perf_bisect
  Benchmark    : system_health.memory_mobile
  Metric       : memory:chrome:all_processes:reported_by_os:system_memory:proportional_resident_size_avg/blank_about/blank_about_blank

Revision             Result                    N
chromium@447645      102598440 +- 3474955      21      good
chromium@447696      103060459 +- 6568885      21      bad

Please refer to the following doc on diagnosing memory regressions:
  https://chromium.googlesource.com/chromium/src/+/master/docs/memory-infra/memory_benchmarks.md

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=blank.about.blank system_health.memory_mobile

Debug Info
  https://chromeperf.appspot.com/buildbucket_job_status/8988183214898376144

Is this bisect wrong?
  https://chromeperf.appspot.com/bad_bisect?try_job_id=6160516088594432


| 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 Tests>AutoBisect.  Thank you!
Project Member

Comment 5 by 42576172...@developer.gserviceaccount.com, Feb 11 2017

Cc: kjellander@chromium.org
Owner: kjellander@chromium.org

=== Auto-CCing suspected CL author kjellander@chromium.org ===

Hi kjellander@chromium.org, the bisect results pointed to your CL, please take a look at the
results.


=== BISECT JOB RESULTS ===
Perf regression found with culprit

Suspected Commit
  Author : kjellander
  Commit : 529ccd92e2a916c7e5650a7222ca09b4406b97da
  Date   : Thu Feb 02 12:09:53 2017
  Subject: Update WebRTC video quality test to use statically linked zxing

Bisect Details
  Configuration: android_nexus5X_perf_bisect
  Benchmark    : system_health.memory_mobile
  Metric       : memory:chrome:all_processes:reported_by_os:system_memory:proportional_resident_size_avg/blank_about/blank_about_blank
  Change       : 0.39% | 101965169.143 -> 101570344.0

Revision             Result                    N
chromium@447537      101965169 +- 1237633      14      good
chromium@447641      102662440 +- 689548       6       good
chromium@447693      103426173 +- 199502       6       good
chromium@447719      104937597 +- 5759925      6       good
chromium@447732      105217320 +- 5598907      6       good
chromium@447739      104230013 +- 267363       6       good
chromium@447742      103161469 +- 2134174      6       good
chromium@447744      102504573 +- 339739       6       good
chromium@447745      101570344 +- 527488       9       bad       <--

Please refer to the following doc on diagnosing memory regressions:
  https://chromium.googlesource.com/chromium/src/+/master/docs/memory-infra/memory_benchmarks.md

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=blank.about.blank system_health.memory_mobile

Debug Info
  https://chromeperf.appspot.com/buildbucket_job_status/8988006078149185792

Is this bisect wrong?
  https://chromeperf.appspot.com/bad_bisect?try_job_id=5294170148700160


| 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!
Cc: primiano@chromium.org hjd@chromium.org perezju@chromium.org
Owner: ----
+hjd, primiano, perezju: The bisect in #5 shows a pattern really similar to https://bugs.chromium.org/p/chromium/issues/detail?id=688448#c19

Removing kjellander because it's extremely unlikely his CL is the culprit, even if bisects continue to pinpoint it for the same reason as in the other bug.
Cc: -kjellander@chromium.org
Right, my CL only affects WebRTC video quality tests running in https://build.chromium.org/p/chromium.webrtc/waterfall and https://build.chromium.org/p/chromium.webrtc.fyi/waterfall - and they only run for the desktop platform.

Comment 8 by hjd@chromium.org, Feb 13 2017

Owner: hjd@chromium.org
Status: Assigned (was: Untriaged)
Yep it looks like the same bug although this time it's hiding a real regression, (although it kind of looks like the all_processes:reported_by_os:system_memory:proportional_resident_size_avg is still going up in the graph? [1] in which case maybe it's not real and a bug.)

We can track the bisect bug at crbug.com/690042 and the regression here. One option to get this bisect to work right now would be to temporarily change the bot to restart the phone between runs.

[1]: https://chromeperf.appspot.com/report?sid=b3f7d52a78eac8098f599c2dd7342e92a9d5c04347b75b03ad51db1bd0201026&start_rev=422373&end_rev=449906

Cc: benhenry@chromium.org
Summary: 1.4MB regression in system_health.memory_mobile at 447218:449231 (was: 1% regression in system_health.memory_mobile at 447646:447696)
There seems to be indeed some 1.4MB regression in overall PSS but, as usual, it's complicated ...

I'm widening the range from 447218 (when metrics were stable last time) to 449231 (when metrics appear to be stable again ... at least for a while).

Breaking up by processes we see:
https://chromeperf.appspot.com/report?sid=d27271b96ab7cc8fce7063dafec83e4085525977f926b3c3cf8e7dde186a69c5&rev=447696

On the browser process:
- 447578:447645 +0.8MB
- 447931:447964 -0.8MB
- 448048:448118 +1.0MB
- 448406:448453 -0.5MB
- Total: +0.5MB

On the renderer processes (ignoring up-downs that cancel each other):
- 447746:447776 +0.3MB
- 448223:448238 +0.2MB
- 448603:448630 +0.2MB
- Total: +0.7MB

On the gpu process:
- 448406:448453 +0.2MB
- Total: +0.2MB

Grand Total: ~ +1.4MB

So many small events adding up to a larger regression :(

Not sure what would be the best way to proceed?
Should we try to track down those two ranges in Browser that are large? It doesn't look like a 1.4MB regression in overall is a dealbreaker with Android, but this is not great.
Started a bisect on the second large bump in the browser process. Let's see if we find something there.
Project Member

Comment 13 by 42576172...@developer.gserviceaccount.com, Feb 16 2017


=== BISECT JOB RESULTS ===
NO Perf regression found

Bisect Details
  Configuration: android_nexus5X_perf_bisect
  Benchmark    : system_health.memory_mobile
  Metric       : memory:chrome:browser_process:reported_by_os:proportional_resident_size_avg/blank_about/blank_about_blank
  Change       : 0.23% | 94271949.7143 -> 94489135.2381

Revision             Result                   N
chromium@448047      94271950 +- 2160809      21      good
chromium@448118      94489135 +- 2809591      21      bad

Please refer to the following doc on diagnosing memory regressions:
  https://chromium.googlesource.com/chromium/src/+/master/docs/memory-infra/memory_benchmarks.md

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=blank.about.blank system_health.memory_mobile

Debug Info
  https://chromeperf.appspot.com/buildbucket_job_status/None

Is this bisect wrong?
  https://chromeperf.appspot.com/bad_bisect?try_job_id=6096717805518848


| 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!
Project Member

Comment 14 by 42576172...@developer.gserviceaccount.com, Feb 16 2017


=== BISECT JOB RESULTS ===
NO Perf regression found

Bisect Details
  Configuration: android_nexus5X_perf_bisect
  Benchmark    : system_health.memory_mobile
  Metric       : memory:chrome:browser_process:reported_by_os:proportional_resident_size_avg/blank_about/blank_about_blank

Revision             Result                   N
chromium@448047      94292674 +- 1929812      21      good
chromium@448118      94490306 +- 2830665      21      bad

Please refer to the following doc on diagnosing memory regressions:
  https://chromium.googlesource.com/chromium/src/+/master/docs/memory-infra/memory_benchmarks.md

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=blank.about.blank system_health.memory_mobile

Debug Info
  https://chromeperf.appspot.com/buildbucket_job_status/8987510755936544624

Is this bisect wrong?
  https://chromeperf.appspot.com/bad_bisect?try_job_id=6684908946194432


| 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!
Labels: Performance-Memory
Status: WontFix (was: Assigned)
This is 6 months old and the bisects didn't reproduce, closing.

Sign in to add a comment