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

Issue 677869 link

Starred by 1 user

Issue metadata

Status: Untriaged
Owner: ----
Cc:
Components:
EstimatedDays: ----
NextAction: ----
OS: ----
Pri: 2
Type: Bug



Sign in to add a comment

Perf JSON output for chromium.perf.fyi/mac-test-retina/battor.power_cases is absolutely massive, causing benchmark failures

Project Member Reported by charliea@chromium.org, Jan 3 2017

Issue description

Strangely, this isn't happening on the main chromium.perf waterfall on the exact same hardware configuration.

The perf JSON output contains lots of histograms that I wouldn't expect. Here's an example piece of output:

"""
"http://edition.cnn.com": {"std": 0.0, "name": "ForInPrepare:count_min", "page_id": 13, "improvement_direction": "down", "important": false, "values": [113], "units": "count", "type": "list_of_scalar_values"}, "http://www.msn.com/ar-ae": {"std": 0.0, "name": "ForInPrepare:count_min", "page_id": 17, "improvement_direction": "down", "important": false, "values": [23], "units": "count", "type": "list_of_scalar_values"}, "https://www.youtube.com/watch?v=_kZsOISarzg": {"std": 0.0, "name": "ForInPrepare:count_min", "page_id": 2, "improvement_direction": "down", "important": false, "values": [7], "units": "count", "type": "list_of_scalar_values"}, "http://www.twitter.com/taylorswift13": {"std": 0.0, "name": "ForInPrepare:count_min", "page_id": 10, "improvement_direction": "down", "important": false, "values": [64], "units": "count", "type": "list_of_scalar_values"}, "http://pollouer.muc/Speedometer/CustomRunner.html?backbone": {"std": 0.0, "name": "ForInPrepare:count_min", "page_id": 31, "improvement_direction": "down", "important": false, "values": [8], "units": "count", "type": "list_of_scalar_values"}, "http://www.sina.com.cn": {"std": 0.0, "name": "ForInPrepare:count_min", "page_id": 20, "improvement_direction": "down", "important": false, "values": [13], "units": "count", "type": "list_of_scalar_values"}, "summary": {"std": 0.0, "name": "ForInPrepare:count_min", "improvement_direction": "down", "important": false, "values": [4, 4, 7, 10, 18, 145, 559, 21, 64, 24, 51, 113, 97, 25, 8, 23, 26, 13, 14, 8, 28, 5, 26, 16, 1, 8, 8, 4], "units": "count", "type": "list_of_scalar_values"}, "https://www.linkedin.com/m/": {"std": 0.0, "name": "ForInPrepare:count_min", "page_id": 16, "improvement_direction": "down", "important": false, "values": [8], "units": "count", "type": "list_of_scalar_values"}, "https://www.google.de/search?q=v8": {"std": 0.0, "name": "ForInPrepare:count_min", "page_id": 0, "improvement_direction": "down", "important": false, "values": [4], "units": "count", "type": "list_of_scalar_values"}, "http://pollouer.muc/Speedometer/CustomRunner.html?jquery": {"std": 0.0, "name": "ForInPrepare:count_min", "page_id": 30, "improvement_direction": "down", "important": false, "values": [8], "units": "count", "type": "list_of_scalar_values"}}, "API_ObjectTemplate_NewInstance:duration_sum": {"http://www.msn.com/ar-ae": {"std": 0.0, "name": "API_ObjectTemplate_NewInstance:duration_sum", "page_id": 17, "improvement_direction": "down", "important": false, "values": [0.224], "units": "ms", "type": "list_of_scalar_values"}, "http://www.twitter.com/taylorswift13": {"std": 0.0, "name": "API_ObjectTemplate_NewInstance:duration_sum", "page_id": 10, "improvement_direction": "down", "important": false, "values": [0.042], "units": "ms", "type": "list_of_scalar_values"}, "http://www.pinterest.com/categories/popular": {"std": 0.0, "name": "API_ObjectTemplate_NewInstance:duration_sum", "page_id": 19, "improvement_direction": "down", "important": false, "values": [0.044], "units": "ms", "type": "list_of_scalar_values"},
"""

What?! It looks like it's creating metrics for every trace event name. I'm not at all sure why it would be doing this.
 
Cc: eakuefner@chromium.org benjhayden@chromium.org
+Ethan, Ben, was one of you planning on testing how much JSON we can load? (Note: I think Charlie's planned fix is to upload less, but it's still another good test case)
Yes, these issues are top-of-mind for me personally at this time. Annie, thanks for bringing this to my attention as another good test case. Charlie, can you link an example build that produces this amount of output?

I'll figure out what the bug tracking situation is on the Catapult side and revisit this one with my findings, but my feeling is that all these cases could benefit from some measurement and optimization on the add_point side.
I've filed https://github.com/catapult-project/catapult/issues/3108 to track this general issue on the Catapult side and cross-referenced this issue from that one.
Charlie, you haven't made any change here yet, right? Which means any build from https://build.chromium.org/p/chromium.perf.fyi/builders/Mac%20Test%20Retina%20Perf should still be fair game for grabbing JSON to investigate this issue from the dashboard side?
Cc: -eakuefner@chromium.org
Components: Test>Telemetry
Components: -Tests>Telemetry

Sign in to add a comment