All graphs for this bug:
  https://chromeperf.appspot.com/group_report?bug_id=656744

Original alerts at time of bug-filing:
  https://chromeperf.appspot.com/group_report?keys=agxzfmNocm9tZXBlcmZyFAsSB0Fub21hbHkYgICg7ejnvgoM


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

android-nexus6

Started bisect job https://chromeperf.appspot.com/buildbucket_job_status/8998530365872594512

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


===== TESTED REVISIONS =====
Revision         Mean     Std Dev  N  Good?
chromium@423027  24715.2  18.9526  5  good
chromium@423062  24725.9  18.427   8  good
chromium@423080  24743.6  14.212   8  good
chromium@423097  24765.1  16.3221  8  bad

Bisect job ran on: android_nexus6_perf_bisect
Bug ID: 656744

Test Command: src/tools/perf/run_benchmark -v --browser=android-chromium --output-format=chartjson --upload-results --pageset-repeat=1 --also-run-disabled-tests storage.indexeddb_endure
Test Metric: vm_private_dirty_final_browser/vm_private_dirty_final_browser
Relative Change: 0.20%
Score: 0

Buildbot stdio: http://build.chromium.org/p/tryserver.chromium.perf/builders/android_nexus6_perf_bisect/builds/2652
Job details: https://chromeperf.appspot.com/buildbucket_job_status/8998530365872594512


Not what you expected? We'll investigate and get back to you!
  https://chromeperf.appspot.com/bad_bisect?try_job_id=5804919112597504

| 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!

Started bisect job https://chromeperf.appspot.com/buildbucket_job_status/8998445040453955856

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


=== Bisection aborted ===
The bisect was aborted because The metric values for the initial "good" and "bad" revisions do not represent a clear regression.
Please contact the the team (see below) if you believe this is in error.

=== Warnings ===
The following warnings were raised by the bisect job:

 * Bisect failed to reproduce the regression with enough confidence.

===== TESTED REVISIONS =====
Revision         Mean     Std Dev  N   Good?
chromium@423080  24700.1  22.185   18  good
chromium@423097  24706.9  19.836   18  bad

Bisect job ran on: android_nexus6_perf_bisect
Bug ID: 656744

Test Command: src/tools/perf/run_benchmark -v --browser=android-chromium --output-format=chartjson --upload-results --pageset-repeat=1 --also-run-disabled-tests storage.indexeddb_endure
Test Metric: vm_private_dirty_final_browser/vm_private_dirty_final_browser
Relative Change: 0.04%
Score: 0

Buildbot stdio: http://build.chromium.org/p/tryserver.chromium.perf/builders/android_nexus6_perf_bisect/builds/2657
Job details: https://chromeperf.appspot.com/buildbucket_job_status/8998445040453955856


Not what you expected? We'll investigate and get back to you!
  https://chromeperf.appspot.com/bad_bisect?try_job_id=5796716094160896

| 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!

Started bisect job https://chromeperf.appspot.com/buildbucket_job_status/8998431873673084336

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


=== Bisection aborted ===
The bisect was aborted because The metric values for the initial "good" and "bad" revisions do not represent a clear regression.
Please contact the the team (see below) if you believe this is in error.

=== Warnings ===
The following warnings were raised by the bisect job:

 * Bisect failed to reproduce the regression with enough confidence.

===== TESTED REVISIONS =====
Revision         Mean     Std Dev  N   Good?
chromium@423080  24725.6  30.3085  18  good
chromium@423097  24722.7  20.1378  18  bad

Bisect job ran on: android_nexus6_perf_bisect
Bug ID: 656744

Test Command: src/tools/perf/run_benchmark -v --browser=android-chromium --output-format=chartjson --upload-results --pageset-repeat=1 --also-run-disabled-tests storage.indexeddb_endure
Test Metric: vm_private_dirty_final_browser/vm_private_dirty_final_browser
Relative Change: 0.03%
Score: 0

Buildbot stdio: http://build.chromium.org/p/tryserver.chromium.perf/builders/android_nexus6_perf_bisect/builds/2663
Job details: https://chromeperf.appspot.com/buildbucket_job_status/8998431873673084336


Not what you expected? We'll investigate and get back to you!
  https://chromeperf.appspot.com/bad_bisect?try_job_id=5319774102355968

| 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!

Started bisect job https://chromeperf.appspot.com/buildbucket_job_status/8998342587245078112

Started bisect job https://chromeperf.appspot.com/buildbucket_job_status/8998342565169174480

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

Hi ryansturm@chromium.org, the bisect results pointed to your CL below as possibly
causing a regression. Please have a look at this info and see whether
your CL be related.


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


===== SUSPECTED CL(s) =====
Subject : Adding a SQL implementation of the backing store for previews opt outs
Author  : ryansturm
Commit description:
  
Previews maintains opt outs to determine eligibility of a navigation to
a domain. To persist this information, a SQLite database is created to
store opt out choice, host name, preview type, and time. This portion of
the implementation only implements creating the table, and returning an
empty result to PreviewBlackList (after running a real query).

BUG= 653730 

Review-Url: https://codereview.chromium.org/2390773003
Cr-Commit-Position: refs/heads/master@{#424319}
Commit  : 2ae78ccf536394379900253a4099b5e089b07084
Date    : Tue Oct 11 01:15:30 2016


===== TESTED REVISIONS =====
Revision         Mean     Std Dev  N  Good?
chromium@423027  24684.4  19.578   5  good
chromium@423785  24692.4  15.5338  5  good
chromium@424164  24719.0  24.8193  5  good
chromium@424259  24764.0  16.4165  5  good
chromium@424306  24748.8  14.4118  5  good
chromium@424318  24736.4  21.3612  5  good
chromium@424319  24832.4  19.4628  5  bad    <--
chromium@424320  24824.8  15.3525  5  bad
chromium@424321  24818.8  13.6272  5  bad
chromium@424324  24813.6  23.1689  5  bad
chromium@424330  24807.4  22.1427  5  bad
chromium@424353  24852.4  10.8074  5  bad
chromium@424542  24792.2  21.2179  5  bad

Bisect job ran on: android_nexus6_perf_bisect
Bug ID: 656744

Test Command: src/tools/perf/run_benchmark -v --browser=android-chromium --output-format=chartjson --upload-results --pageset-repeat=1 --also-run-disabled-tests storage.indexeddb_endure
Test Metric: vm_private_dirty_final_browser/vm_private_dirty_final_browser
Relative Change: 0.44%
Score: 99.9

Buildbot stdio: http://build.chromium.org/p/tryserver.chromium.perf/builders/android_nexus6_perf_bisect/builds/2674
Job details: https://chromeperf.appspot.com/buildbucket_job_status/8998342565169174480


Not what you expected? We'll investigate and get back to you!
  https://chromeperf.appspot.com/bad_bisect?try_job_id=5827271565246464

| 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!

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

Hi dominickn@chromium.org, the bisect results pointed to your CL below as possibly
causing a regression. Please have a look at this info and see whether
your CL be related.


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


===== SUSPECTED CL(s) =====
Subject : [Reland] Refactor WebappRegistry into a singleton instance.
Author  : dominickn
Commit description:
  
This CL refactors WebappRegistry and WebappDataStorage to make most of
the methods synchronous. WebappRegistry is now a singleton instance that
is instantiated at browser startup. This allows all SharedPreferences files to
be pre-warmed before the class is used; new web apps open new
SharedPreferences on a background thread when registered, after which the
preferences are cached automatically.

Most static methods on WebappRegistry and WebappDataStorage have been
converted to instance methods or removed. This makes the code much
cleaner and more efficient; each static method had to independently open
their SharedPreferences, which minimally performs a stat() on the
underlying XML file to see if it has changed. Now the singleton
WebappRegistry caches all WebappDataStorage objects on startup and
whenever new ones are added. This reduces disk IO overhead.

This CL allows all calls to SharedPreferences.Editor.apply() in
WebappRegistry and WebappDataStorage to occur on the main thread,
mostly removing the need for unwieldy callback interfaces and bare
pointer passing across the JNI.

BUG= 633791 

Committed: https://crrev.com/78cf6596040662c0a73c0e4038bc84fb2b3c470a
Review-Url: https://codereview.chromium.org/2351113005
Cr-Original-Commit-Position: refs/heads/master@{#422703}
Cr-Commit-Position: refs/heads/master@{#423077}
Commit  : 6e08a0b7993fb106e3dea796950521cfd04a47cf
Date    : Wed Oct 05 04:55:47 2016


===== TESTED REVISIONS =====
Revision         Mean     Std Dev  N   Good?
chromium@423027  24699.8  21.5913  18  good
chromium@423062  24712.1  18.8912  18  good
chromium@423071  24704.4  14.9257  9   good
chromium@423076  24707.5  20.368   8   good
chromium@423077  24735.6  4.9295   5   bad    <--
chromium@423078  24736.5  11.7473  8   bad
chromium@423080  24739.1  18.5583  8   bad
chromium@423097  24730.3  21.5237  18  bad

Bisect job ran on: android_nexus6_perf_bisect
Bug ID: 656744

Test Command: src/tools/perf/run_benchmark -v --browser=android-chromium --output-format=chartjson --upload-results --pageset-repeat=1 --also-run-disabled-tests storage.indexeddb_endure
Test Metric: vm_private_dirty_final_browser/vm_private_dirty_final_browser
Relative Change: 0.14%
Score: 99.0

Buildbot stdio: http://build.chromium.org/p/tryserver.chromium.perf/builders/android_nexus6_perf_bisect/builds/2673
Job details: https://chromeperf.appspot.com/buildbucket_job_status/8998342587245078112


Not what you expected? We'll investigate and get back to you!
  https://chromeperf.appspot.com/bad_bisect?try_job_id=5870563124314112

| 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!

The suspected iteration of the CL was reverted in chromium@423138. It landed for good in chromium@425214 after further revisions.

Also, does this test mean include startup time? The only difference my CL should make on performance tests that don't touch PWA UI related code is that startup performance may be slightly more variable.

+sullivan@ -- I don't see any owners for this test, so I'm not sure who to nag for additional context. Recommended next steps?

I'm not really sure how my CL could affect IndexedDB. My CL only has code that runs at profile start up and tear down, since the SQLite DB is not written to yet. I'm not really sure what this perf test measures, and knowing more about it would be helpful.

+dmurph, who wrote these tests. Daniel, note that the regressions here are very very small: bisect reproduced it only in 2 of 5 runs, and in those runs a 0.14% change and a 0.44% change. The perf dashboard only saw a 1% regression. So the first question is if we care about a regression that is so small. Feel free to WontFix if we do not care.

If we do care, how should ryansturm investigate? Is it likely that his CL could cause this?

Is this memory usage that's regressing?

Yes, it is memory usage, you can see buried in the bisect comments:

"Test Metric: vm_private_dirty_final_browser/vm_private_dirty_final_browser"

IndexedDB doesn't used SQLLite, it uses LevelDB. I'm adding Chris just to make sure that there's nothing hidden happening here

The only thing IDB related seems to be https://codereview.chromium.org/2366253002 "Remove ActiveDOMObject::stop()". However, this test has historically been quite noisy, and looking at the prior history of this graph, memory usage goes up and down quite a lot.

Per #19, I'm closing this. Noisy benchmark, small graph movement, and no repro.

Started bisect job https://chromeperf.appspot.com/buildbucket_job_status/8982580245539143888