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

Issue 667326 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

5.1%-15.1% regression in tracing.tracing_with_debug_overhead at 433427:433451

Project Member Reported by mustaq@chromium.org, Nov 21 2016

Issue description

See the link to graphs below.
 
Project Member

Comment 3 by 42576172...@developer.gserviceaccount.com, Nov 21 2016


===== BISECT JOB RESULTS =====
Status: failed


=== Bisection aborted ===
The bisect was aborted because Bisect cannot identify a culprit: Bisect failed to reproduce the regression with enough confidence.
Please contact the the team (see below) if you believe this is in error.

===== TESTED REVISIONS =====
Revision         Mean     Std Dev  N   Good?
chromium@433427  30577.5  136328   45  good
chromium@433437  30967.7  144550   45  bad

Bisect job ran on: android_nexus9_perf_bisect
Bug ID: 667326

Test Command: src/tools/perf/run_benchmark -v --browser=android-chromium --output-format=chartjson --upload-results --pageset-repeat=1 --also-run-disabled-tests tracing.tracing_with_debug_overhead
Test Metric: Max number of events per second_min/Max number of events per second_min
Relative Change: None

Buildbot stdio: http://build.chromium.org/p/tryserver.chromium.perf/builders/android_nexus9_perf_bisect/builds/2273
Job details: https://chromeperf.appspot.com/buildbucket_job_status/8995368823296436400


Not what you expected? We'll investigate and get back to you!
  https://chromeperf.appspot.com/bad_bisect?try_job_id=5822279097778176

| 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 6 by 42576172...@developer.gserviceaccount.com, Nov 21 2016


===== BISECT JOB RESULTS =====
Status: failed


=== Bisection aborted ===
The bisect was aborted because Bisect cannot identify a culprit: Bisect failed to reproduce the regression with enough confidence.
Please contact the the team (see below) if you believe this is in error.

===== TESTED REVISIONS =====
Revision         Mean     Std Dev  N   Good?
chromium@433427  27772.0  134120   45  good
chromium@433451  30291.8  141552   45  bad

Bisect job ran on: android_nexus9_perf_bisect
Bug ID: 667326

Test Command: src/tools/perf/run_benchmark -v --browser=android-chromium --output-format=chartjson --upload-results --pageset-repeat=1 --also-run-disabled-tests tracing.tracing_with_debug_overhead
Test Metric: Max number of events per second_max/Max number of events per second_max
Relative Change: None

Buildbot stdio: http://build.chromium.org/p/tryserver.chromium.perf/builders/android_nexus9_perf_bisect/builds/2275
Job details: https://chromeperf.appspot.com/buildbucket_job_status/8995363014566084096


Not what you expected? We'll investigate and get back to you!
  https://chromeperf.appspot.com/bad_bisect?try_job_id=5779554105294848

| 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 7 by 42576172...@developer.gserviceaccount.com, Nov 21 2016


===== BISECT JOB RESULTS =====
Status: failed


=== Bisection aborted ===
The bisect was aborted because Bisect cannot identify a culprit: Bisect failed to reproduce the regression with enough confidence.
Please contact the the team (see below) if you believe this is in error.

===== TESTED REVISIONS =====
Revision         Mean     Std Dev  N   Good?
chromium@433427  27911.5  134492   45  good
chromium@433451  29065.6  139231   45  bad

Bisect job ran on: android_nexus9_perf_bisect
Bug ID: 667326

Test Command: src/tools/perf/run_benchmark -v --browser=android-chromium --output-format=chartjson --upload-results --pageset-repeat=1 --also-run-disabled-tests tracing.tracing_with_debug_overhead
Test Metric: Max number of events per second_avg/Max number of events per second_avg
Relative Change: None

Buildbot stdio: http://build.chromium.org/p/tryserver.chromium.perf/builders/android_nexus9_perf_bisect/builds/2274
Job details: https://chromeperf.appspot.com/buildbucket_job_status/8995363029239150544


Not what you expected? We'll investigate and get back to you!
  https://chromeperf.appspot.com/bad_bisect?try_job_id=5799773569810432

| 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 9 by 42576172...@developer.gserviceaccount.com, Nov 22 2016


===== BISECT JOB RESULTS =====
Status: failed


=== Bisection aborted ===
The bisect was aborted because Bisect cannot identify a culprit: Bisect failed to reproduce the regression with enough confidence.
Please contact the the team (see below) if you believe this is in error.

===== TESTED REVISIONS =====
Revision         Mean     Std Dev  N   Good?
chromium@433427  27208.5  130701   45  good
chromium@433451  30123.2  140137   45  bad

Bisect job ran on: android_nexus9_perf_bisect
Bug ID: 667326

Test Command: src/tools/perf/run_benchmark -v --browser=android-chromium --output-format=chartjson --upload-results --pageset-repeat=1 --also-run-disabled-tests tracing.tracing_with_debug_overhead
Test Metric: Max number of events per second_sum/Max number of events per second_sum
Relative Change: None

Buildbot stdio: http://build.chromium.org/p/tryserver.chromium.perf/builders/android_nexus9_perf_bisect/builds/2280
Job details: https://chromeperf.appspot.com/buildbucket_job_status/8995271967651721824


Not what you expected? We'll investigate and get back to you!
  https://chromeperf.appspot.com/bad_bisect?try_job_id=5622732501286912

| 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 11 by 42576172...@developer.gserviceaccount.com, Nov 29 2016


===== BISECT JOB RESULTS =====
Status: completed


===== TESTED REVISIONS =====
Revision         Mean     Std Dev  N   Good?
chromium@433427  22791.7  15947.8  18  good
chromium@433433  23385.1  5924.17  18  good
chromium@433436  22792.2  13835.1  18  good
chromium@433437  24054.1  20877.6  27  unknown
chromium@433438  24382.3  18135.6  27  unknown
chromium@433439  25035.9  15635.4  18  bad
chromium@433451  25847.7  10213.8  12  bad

Bisect job ran on: android_nexus9_perf_bisect
Bug ID: 667326

Test Command: src/tools/perf/run_benchmark -v --browser=android-chromium --output-format=chartjson --upload-results --pageset-repeat=1 --also-run-disabled-tests --story-filter=http...www.ask.com. tracing.tracing_with_debug_overhead
Test Metric: Max number of events per second_avg/http___www.ask.com_
Relative Change: 14.28%

Buildbot stdio: http://build.chromium.org/p/tryserver.chromium.perf/builders/android_nexus9_perf_bisect/builds/2288
Job details: https://chromeperf.appspot.com/buildbucket_job_status/8994628386916809280


Not what you expected? We'll investigate and get back to you!
  https://chromeperf.appspot.com/bad_bisect?try_job_id=6744210235981824

| 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 15 by 42576172...@developer.gserviceaccount.com, Apr 11 2017


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

Bisect Details
  Configuration: android_nexus9_perf_bisect
  Benchmark    : tracing.tracing_with_debug_overhead
  Metric       : Max number of events per second_sum/Max number of events per second_sum

Revision             Result                 N
chromium@433427      27208.5 +- 130701      45      good
chromium@433451      30123.2 +- 140137      45      bad

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 tracing.tracing_with_debug_overhead

Debug Info
  https://chromeperf.appspot.com/buildbucket_job_status/8982611777959419968

Is this bisect wrong?
  https://chromeperf.appspot.com/bad_bisect?try_job_id=5622732501286912


| 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 16 by 42576172...@developer.gserviceaccount.com, Apr 11 2017


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

Bisect Details
  Configuration: android_nexus9_perf_bisect
  Benchmark    : tracing.tracing_with_debug_overhead
  Metric       : Max number of events per second_max/Max number of events per second_max

Revision             Result                 N
chromium@433427      27772.0 +- 134120      45      good
chromium@433451      30291.8 +- 141552      45      bad

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 tracing.tracing_with_debug_overhead

Debug Info
  https://chromeperf.appspot.com/buildbucket_job_status/8982592590297140304

Is this bisect wrong?
  https://chromeperf.appspot.com/bad_bisect?try_job_id=5779554105294848


| 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 17 by 42576172...@developer.gserviceaccount.com, Apr 11 2017


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

Bisect Details
  Configuration: android_nexus9_perf_bisect
  Benchmark    : tracing.tracing_with_debug_overhead
  Metric       : Max number of events per second_avg/Max number of events per second_avg

Revision             Result                 N
chromium@433427      27911.5 +- 134492      45      good
chromium@433451      29065.6 +- 139231      45      bad

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 tracing.tracing_with_debug_overhead

Debug Info
  https://chromeperf.appspot.com/buildbucket_job_status/8982580424068338064

Is this bisect wrong?
  https://chromeperf.appspot.com/bad_bisect?try_job_id=5799773569810432


| 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: zh...@chromium.org oysteine@chromium.org nednguyen@chromium.org
cc-ing the test owners re what to do here:

Most of the 28 regressions here are valid but in the last 3+ months many of them has seen even bigger regressions later on.

Do we care about the ones in this bug?
Zhen, Oysteine: I think we need to declare that the tracing benchmarks is a failure: it keeps producing unbisectable regressions & the breakdown data is not good enough for us to address the cause. Given the fact that this is not producing positive value in term of regression prevention, I think we should turn down the benchmarks & retry a better approach once we have more bandwidth. Wdyt?
Cc: -nednguyen@chromium.org
Owner: nednguyen@chromium.org
Status: Assigned (was: Untriaged)
Cc: -zh...@chromium.org -mustaq@chromium.org nedngu...@google.com
Owner: zh...@chromium.org
Reassign to Zhen & Oysteine to answer to #19

Comment 22 by zh...@chromium.org, Apr 12 2017

Is that because the trace size depends on how long it takes to run a story, which could vary from run to run?
Unclear, but IIRC, there are still some bug in the current implementation of max number events per second metric that make it incorrect in the case of clock sync.

Comment 24 by zh...@chromium.org, Apr 12 2017

What kind of bug is it?

Oystein, have you guys found the tracing benchmark useful to tracing? I haven't followed that thread much.
Zhen: it was https://bugs.chromium.org/p/chromium/issues/detail?id=646521#c22

My CL was adding around 200 trace events at total to the combined trace, but it regressed the metric by 34% (number of Chrome's trace events is tens of thousands)

Comment 26 by zh...@chromium.org, May 15 2017

Cc: charliea@chromium.org
I just chatted with Oystein offline. Can we turn off the alert for this until either of us has more time digging into it more?
#26: Why not just turn off the benchmark entirely then? Otherwise I think we are just wasting bot time if we don't have bandwidth to address its issues :-)

Comment 28 by zh...@chromium.org, May 15 2017

Oystein, do we need to keep the benchmark running? I guess any new trace event additions can use this benchmark locally for manual verification. If not, I can prepare a disable CL.
For locally run benchmark, you can move it to "tools/perf/contrib/". Benchmarks there can be run locally & on tryserver. THey are also not scheduled on perf waterfall, so there is no need to add @Decorators.Disabled('...').
Do we still want this benchmark? If not, who can move it to contrib?
Ping oysteine@: I would be ok with moving this to tools/perf/contrib/

Comment 32 by zh...@chromium.org, Aug 17 2017

I am fine with the move. Oystein?
Yep that sounds reasonable, since no one is really in a position to maintain this at the moment.

Comment 35 by zh...@chromium.org, Aug 18 2017

Status: WontFix (was: Assigned)

Sign in to add a comment