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

Issue 848551 link

Starred by 2 users

Issue metadata

Status: Fixed
Owner:
Closed: Jun 2018
Cc:
Components:
EstimatedDays: ----
NextAction: ----
OS: Linux , Windows , Mac
Pri: 2
Type: Bug-Regression
Proj-Servicification

Blocking:
issue 824783



Sign in to add a comment

Pinpoint bisection for chromium.perf.fyi mojo-linux-perf

Project Member Reported by chongz@chromium.org, Jun 1 2018

Issue description

All graphs for this bug:
  https://chromeperf.appspot.com/group_report?bug_id=848551

(For debugging:) Original alerts at time of bug-filing:
  https://chromeperf.appspot.com/group_report?sid=e12823e736504c48330606c8f3c073c5f447e79353fba8c5b476b980e8cf1db6


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

mojo-linux-perf
Blocking: 824783
Cc: nednguyen@chromium.org
Components: Speed>Bisection Internals>Services>Network
Summary: Bisect loading.desktop.network_service failed - Error: Builder linux_perf_bisect not found (was: 12180.1% improvement in loading.desktop.network_service at 560176:560186)
The job above returns the following error:
```
DATA:{
    "bucket": "master.tryserver.chromium.perf", 
    "canary_preference": "AUTO", 
    "completed_utc": "Fri Jun  1 01:45:21 2018", 
    "created_by": "user:425761728072-pa1bs18esuhp2cp2qfa1u9vb6p1v6kfu@developer.gserviceaccount.com", 
    "created_utc": "Fri Jun  1 01:45:14 2018", 
    "experimental": false, 
    "failure_reason": "INVALID_BUILD_DEFINITION", 
    "id": "8944960937681560640", 
    "parameters": {
        "builder_name": "linux_perf_bisect", 
        "properties": {
            "bisect_config": {
                "bad_revision": "560196", 
                "bug_id": 848551, 
                "builder_host": null, 
                "builder_port": null, 
                "command": "src/tools/perf/run_benchmark -v --browser=release --output-format=chartjson --upload-results --pageset-repeat=1 --also-run-disabled-tests --story-filter=Flickr.diff loading.desktop.network_service", 
                "good_revision": "560155", 
                "gs_bucket": "chrome-perf", 
                "max_time_minutes": "20", 
                "metric": "timeToFirstContentfulPaint:style_avg_diff/warm/Flickr_diff", 
                "recipe_tester_name": "linux_perf_bisect", 
                "repeat_count": "20", 
                "test_type": "perf", 
                "try_job_id": 5814493810786304
            }
        }
    }, 
    "project": "chromium", 
    "result": "FAILURE", 
    "result_details": {
        "error": {
            "message": "Builder linux_perf_bisect not found"
        }
    }, 
    "status": "COMPLETED", 
    "status_changed_utc": "Fri Jun  1 01:45:21 2018", 
    "tags": [
        "builder:linux_perf_bisect"
    ], 
    "updated_utc": "Fri Jun  1 01:45:21 2018", 
    "utcnow_utc": "Fri Jun  1 20:13:30 2018"
} 
```

Looks like they cannot find the bisect bot (linux_perf_bisect)...

+nednguyen@ Is there any special configurations I need to do to make bisect work?

Thanks!

Cc: dtu@chromium.org
Currently, there is no bisect support for fyi waterfall
Ah I see, thanks for the information!
Do we know how hard it is to add the support? Or are they actually using different infrastructures (other than different priorities)?

Thanks!
It's mostly different priorities. The infrastrucutre is essentially the same

Comment 8 by dxie@google.com, Jun 7 2018

Labels: Hotlist-KnownIssue OS-Linux OS-Mac OS-Windows

Comment 9 by dtu@chromium.org, Jun 7 2018

Cc: simonhatch@chromium.org
To add support, we need to:
* Ensure that the perf builders build that isolate
* Ensure that there are bisector bots matching that FYI bot config, and add some if necessary
* Update some perf dashboard and Pinpoint configs
Re #c9: Thanks for the information! Since I'm not so familiar with the builder config here - Do you have a rough estimate on how many SWE hours required to add the support?

Or if it's a burden for your team, can you point me to related docs or patches so I can take a look and see if I can fix it?

Thanks!

Comment 11 by dtu@chromium.org, Jun 7 2018

Owner: dtu@chromium.org
Status: Assigned (was: Untriaged)
Summary: Pinpoint bisection for chromium.perf.fyi mojo-linux-perf (was: Bisect loading.desktop.network_service failed - Error: Builder linux_perf_bisect not found)

Comment 12 by dtu@chromium.org, Jun 7 2018

II checked, and we should already have that benchmark available on the perf builders. We also have those Linux machines available already, so we just need to update some configs.

Uploaded a CL to add the configs.
https://chromium-review.googlesource.com/c/catapult/+/1091807
Nice, really appreciated!
Project Member

Comment 15 by bugdroid1@chromium.org, Jun 8 2018

The following revision refers to this bug:
  https://chromium.googlesource.com/chromium/src.git/+/a3f039c765a3f71a1341e553518065394179d0ba

commit a3f039c765a3f71a1341e553518065394179d0ba
Author: catapult-chromium-autoroll <catapult-chromium-autoroll@skia-buildbots.google.com.iam.gserviceaccount.com>
Date: Fri Jun 08 04:56:57 2018

Roll src/third_party/catapult 09b82f4..c971895 (1 commits)

https://chromium.googlesource.com/catapult.git/+log/09b82f4..c971895


git log 09b82f4..c971895 --date=short --no-merges --format='%ad %ae %s'
2018-06-07 dtu@chromium.org [dashboard] Add mojo-linux-perf to Pinpoint.


Created with:
  gclient setdep -r src/third_party/catapult@c971895

The AutoRoll server is located here: https://catapult-roll.skia.org

Documentation for the AutoRoller is here:
https://skia.googlesource.com/buildbot/+/master/autoroll/README.md

If the roll is causing failures, please contact the current sheriff, who should
be CC'd on the roll, and stop the roller if necessary.

CQ_INCLUDE_TRYBOTS=luci.chromium.try:android_optional_gpu_tests_rel;luci.chromium.try:linux_optional_gpu_tests_rel;luci.chromium.try:mac_optional_gpu_tests_rel;luci.chromium.try:win_optional_gpu_tests_rel

BUG= chromium:848551 
TBR=sullivan@chromium.org

Change-Id: Iaf08d38a02eb816bcc880cb602ed2592804fdc2d
Reviewed-on: https://chromium-review.googlesource.com/1092250
Reviewed-by: catapult-chromium-autoroll <catapult-chromium-autoroll@skia-buildbots.google.com.iam.gserviceaccount.com>
Commit-Queue: catapult-chromium-autoroll <catapult-chromium-autoroll@skia-buildbots.google.com.iam.gserviceaccount.com>
Cr-Commit-Position: refs/heads/master@{#565543}
[modify] https://crrev.com/a3f039c765a3f71a1341e553518065394179d0ba/DEPS

Comment 16 by dtu@chromium.org, Jun 11 2018

Status: Fixed (was: Assigned)
Configs should be up! Let me know if anything is not working.
Awesome! Will give it a try and get back to you if I still have questions. Thanks!
Project Member

Comment 19 by 42576172...@developer.gserviceaccount.com, Jun 11 2018

😿 Pinpoint job stopped with an error.
https://pinpoint-dot-chromeperf.appspot.com/job/12cb55c5240000

All of the runs failed. The most common error (5/20 runs) was:
  Run at /b/s/w/ir/third_party/catapult/telemetry/telemetry/benchmark_runner.pyTraceback (most recent call last):
Please ignore the failure above - The fix was landed earlier today, will give it some time to propagate.

The job failed because it's trying to use `histograms` as the output format, which is not supported by loading.desktop.network_service:
```From #c21
2621 2018-06-15 00:22:42.664 E: local_auth: the returned access token is invalid
Additional test environment:
    CHROME_DEVEL_SANDBOX=/opt/chromium/chrome_sandbox
    CHROME_HEADLESS=1
    LANG=en_US.UTF-8
Command: /usr/bin/python ../../testing/scripts/run_performance_tests.py ../../tools/perf/run_benchmark --benchmarks loading.desktop.network_service --story-filter Flickr.diff --pageset-repeat 1 --browser release -v --upload-results --output-format histograms --isolated-script-test-output /b/s/w/ioJNlyvN/output.json --isolated-script-test-chartjson-output /b/s/w/ioJNlyvN/chartjson-output.json --results-label chromium@aa324eb
```

However I did changed the output format back to `chartjson` for the benchmark on Jun 11 (bb7c4560ca39360eeb681248378bb7099ff28a00), or do we have a separate config for bisect?

Thanks!
Project Member

Comment 23 by 42576172...@developer.gserviceaccount.com, Jun 15 2018

😿 Pinpoint job stopped with an error.
https://pinpoint-dot-chromeperf.appspot.com/job/13ba8825240000

The swarming task expired. The bots are likely overloaded, dead, or misconfigured.

Comment 24 by dtu@chromium.org, Jun 18 2018

Ah sorry, Pinpoint has already dropped support for chartjson. Pinpoint contains its own logic for reading and interpreting the benchmark results, and supports histograms only.
Ah I see, guess I will have to run pinpoint manually then... Anyway thanks for the help!

There's no manual pinpoint, it's a service that runs on appengine.
Re #c26: Sorry for the confusion. I was thinking I might be able to run pinpoint try jobs with Rev# vs. Rev# + patch, but just find out that we only support HEAD vs. HEAD + patch.

Sign in to add a comment