Issue metadata
Sign in to add a comment
|
No data received for system_health.memory_desktop from chromium-rel-win7-dual since 405235 |
||||||||||||||||||||
Issue descriptionSee the link to graphs below.
,
Aug 8 2016
As shown in https://chromeperf.appspot.com/report?sid=08252a4899c253c74a44a0661974d08f84e01ec732a7ba1e971f80960858b714&start_rev=403915&end_rev=407993, gpu values disappeared in the range r405236-r405763 (http://test-results.appspot.com/revision_range?start=405236&end=405763). I'll try to run a bisect (not sure if this is supported).
,
Aug 8 2016
Started bisect job https://chromeperf.appspot.com/buildbucket_job_status/9004894628517351600
,
Aug 9 2016
Bisect failed: http://build.chromium.org/p/tryserver.chromium.perf/builders/win_perf_bisect/builds/6816 Failure reason: the build has failed. Additional errors: The metric was not found in the test output. Either of the initial "good" or "bad" revisions failed to be tested or built.
,
Aug 9 2016
Started bisect job https://chromeperf.appspot.com/buildbucket_job_status/9004829888067601408
,
Aug 9 2016
Bisect failed: http://build.chromium.org/p/tryserver.chromium.perf/builders/win_perf_bisect/builds/6816 Failure reason: the build has failed. Additional errors: The metric was not found in the test output. Either of the initial "good" or "bad" revisions failed to be tested or built.
,
Aug 15 2016
Started bisect job https://chromeperf.appspot.com/buildbucket_job_status/9004241988482421776
,
Aug 16 2016
Bisect failed: http://build.chromium.org/p/tryserver.chromium.perf/builders/win_perf_bisect/builds/6835 Failure reason: the build has failed. Additional errors: The metric was not found in the test output. Either of the initial "good" or "bad" revisions failed to be tested or built.
,
Aug 16 2016
Started bisect job https://chromeperf.appspot.com/buildbucket_job_status/9004197643524640784
,
Aug 16 2016
Bisect failed: http://build.chromium.org/p/tryserver.chromium.perf/builders/win_perf_bisect/builds/6835 Failure reason: the build has failed. Additional errors: The metric was not found in the test output. Either of the initial "good" or "bad" revisions failed to be tested or built.
,
Aug 16 2016
The bisects are consistently unable to reproduce the issue. More precisely, they don't find the relevant value (memory:chrome:all_processes:reported_by_chrome:gpu:effective_size_avg) even in the good revision (r405235). Given this and the fact that the values had not been reported flakily (before the stoppage), I think that this is due to an infrastructure change. More "interesting" facts: 1. There was almost a two day gap between the last build that reported the value and the next one: Build 10183 end: Wed Jul 13 15:17:37 2016 https://build.chromium.org/p/chromium.perf/builders/Win%207%20Perf%20%283%29/builds/10183/ Build 10184 start: Fri Jul 15 09:51:36 2016 https://build.chromium.org/p/chromium.perf/builders/Win%207%20Perf%20%283%29/builds/10184/ 2. While the trace from last build that reported the value contains the GPU process, the trace from the next build doesn't contain any GPU process tracks at all: Build 10183: https://console.developers.google.com/m/cloudstorage/b/chrome-telemetry-output/o/trace-file-id_21-2016-07-13_14-59-47-84384.html Build 10184: https://console.developers.google.com/m/cloudstorage/b/chrome-telemetry-output/o/trace-file-id_21-2016-07-15_11-46-56-86299.html The traces from the bisect corresponding to the last build reporting the GPU values on the dashboard also don't contain any GPU process tracks: https://console.developers.google.com/m/cloudstorage/b/chrome-telemetry-output/o/trace-file-id_21-2016-08-15_19-38-49-20762.html https://console.developers.google.com/m/cloudstorage/b/chrome-telemetry-output/o/trace-file-id_21-2016-08-15_19-49-57-40271.html https://console.developers.google.com/m/cloudstorage/b/chrome-telemetry-output/o/trace-file-id_21-2016-08-15_20-01-03-39388.html https://console.developers.google.com/m/cloudstorage/b/chrome-telemetry-output/o/trace-file-id_21-2016-08-15_20-12-08-78730.html https://console.developers.google.com/m/cloudstorage/b/chrome-telemetry-output/o/trace-file-id_21-2016-08-15_20-23-11-29029.html I checked the actual model in one of these traces and there is NO GPU PROCESS at all!!! 3. There's a huge drop in the overall effective size (116.0 MiB → 83.2 MiB) between the two builds: https://chromeperf.appspot.com/report?sid=0209a137f7cfb9773e1037f298d78e17246ecf9a8d7f64e01e9090959a7544f2&rev=405235 Annie, Roberto, Dave: Have there been any changes to the chrome-rel-win7-dual builder around July 14th? Eric, David: Any idea why the GPU process would suddenly disappear from the traces? Could a bot configuration change make Chrome do GPU stuff directly from the browser process (as opposed to a separate GPU process)? Could this be caused by a stricter sandboxing policy?
,
Aug 16 2016
Good detective work, Petr! Looks like these bots were replaced on July 15 in bug 611913. Unfortunately logs are unavailable from before the replacement, but here are the logs from after, looks like it is not using GPU. I didn't realize these bots were using GPU previously (I thought only bots with GPU in the name and laptops use GPU). Dave, is that correct? (INFO) 2016-08-16 06:09:19,062 browser._LogBrowserInfo:113 OS: win win7 (INFO) 2016-08-16 06:09:19,062 browser._LogBrowserInfo:120 GPU device 0: VENDOR = 0x102b, DEVICE = 0x532 (INFO) 2016-08-16 06:09:19,062 browser._LogBrowserInfo:122 GPU Attributes: (INFO) 2016-08-16 06:09:19,062 browser._LogBrowserInfo:124 adapter_luid : 0.0 (INFO) 2016-08-16 06:09:19,062 browser._LogBrowserInfo:124 amd_switchable : False (INFO) 2016-08-16 06:09:19,062 browser._LogBrowserInfo:124 basic_info_state : 1 (INFO) 2016-08-16 06:09:19,062 browser._LogBrowserInfo:124 can_lose_context : False (INFO) 2016-08-16 06:09:19,062 browser._LogBrowserInfo:124 context_info_state : 0 (INFO) 2016-08-16 06:09:19,062 browser._LogBrowserInfo:124 direct_rendering : True (INFO) 2016-08-16 06:09:19,062 browser._LogBrowserInfo:124 driver_date : 6-21-2006 (INFO) 2016-08-16 06:09:19,062 browser._LogBrowserInfo:124 driver_vendor : Microsoft (INFO) 2016-08-16 06:09:19,062 browser._LogBrowserInfo:124 driver_version : 6.1.7600.16385 (INFO) 2016-08-16 06:09:19,062 browser._LogBrowserInfo:124 dx_diagnostics_info_state: 0 (INFO) 2016-08-16 06:09:19,062 browser._LogBrowserInfo:124 gl_extensions : (INFO) 2016-08-16 06:09:19,062 browser._LogBrowserInfo:124 gl_renderer : (INFO) 2016-08-16 06:09:19,062 browser._LogBrowserInfo:124 gl_reset_notification_strategy: 0 (INFO) 2016-08-16 06:09:19,062 browser._LogBrowserInfo:124 gl_vendor : (INFO) 2016-08-16 06:09:19,062 browser._LogBrowserInfo:124 gl_version : (INFO) 2016-08-16 06:09:19,062 browser._LogBrowserInfo:124 gl_ws_extensions : (INFO) 2016-08-16 06:09:19,062 browser._LogBrowserInfo:124 gl_ws_vendor : (INFO) 2016-08-16 06:09:19,062 browser._LogBrowserInfo:124 gl_ws_version : (INFO) 2016-08-16 06:09:19,062 browser._LogBrowserInfo:124 in_process_gpu : True (INFO) 2016-08-16 06:09:19,062 browser._LogBrowserInfo:124 initialization_time : 0.0 (INFO) 2016-08-16 06:09:19,062 browser._LogBrowserInfo:124 jpeg_decode_accelerator_supported: False (INFO) 2016-08-16 06:09:19,062 browser._LogBrowserInfo:124 lenovo_dcute : False (INFO) 2016-08-16 06:09:19,062 browser._LogBrowserInfo:124 max_msaa_samples : (INFO) 2016-08-16 06:09:19,062 browser._LogBrowserInfo:124 optimus : False (INFO) 2016-08-16 06:09:19,062 browser._LogBrowserInfo:124 pixel_shader_version: (INFO) 2016-08-16 06:09:19,062 browser._LogBrowserInfo:124 process_crash_count : 0 (INFO) 2016-08-16 06:09:19,062 browser._LogBrowserInfo:124 sandboxed : False (INFO) 2016-08-16 06:09:19,062 browser._LogBrowserInfo:124 software_rendering : False (INFO) 2016-08-16 06:09:19,062 browser._LogBrowserInfo:124 vertex_shader_version: (INFO) 2016-08-16 06:09:19,062 browser._LogBrowserInfo:124 video_decode_accelerator_flags: 0 (INFO) 2016-08-16 06:09:19,062 browser._LogBrowserInfo:126 Feature Status: (INFO) 2016-08-16 06:09:19,062 browser._LogBrowserInfo:128 2d_canvas : unavailable_software (INFO) 2016-08-16 06:09:19,062 browser._LogBrowserInfo:128 flash_3d : unavailable_software (INFO) 2016-08-16 06:09:19,062 browser._LogBrowserInfo:128 flash_stage3d : unavailable_software (INFO) 2016-08-16 06:09:19,062 browser._LogBrowserInfo:128 flash_stage3d_baseline: unavailable_software (INFO) 2016-08-16 06:09:19,062 browser._LogBrowserInfo:128 gpu_compositing : unavailable_software (INFO) 2016-08-16 06:09:19,062 browser._LogBrowserInfo:128 multiple_raster_threads: unavailable_off (INFO) 2016-08-16 06:09:19,062 browser._LogBrowserInfo:128 native_gpu_memory_buffers: disabled_software (INFO) 2016-08-16 06:09:19,062 browser._LogBrowserInfo:128 rasterization : unavailable_software (INFO) 2016-08-16 06:09:19,062 browser._LogBrowserInfo:128 video_decode : unavailable_software (INFO) 2016-08-16 06:09:19,062 browser._LogBrowserInfo:128 video_encode : unavailable_software (INFO) 2016-08-16 06:09:19,062 browser._LogBrowserInfo:128 vpx_decode : unavailable_software (INFO) 2016-08-16 06:09:19,062 browser._LogBrowserInfo:128 webgl : unavailable_off (INFO) 2016-08-16 06:09:19,062 browser._LogBrowserInfo:130 Driver Bug Workarounds: (INFO) 2016-08-16 06:09:19,062 browser._LogBrowserInfo:132 clear_uniforms_before_first_program_use (INFO) 2016-08-16 06:09:19,062 browser._LogBrowserInfo:132 disable_d3d11 (INFO) 2016-08-16 06:09:19,062 browser._LogBrowserInfo:132 disable_direct_composition (INFO) 2016-08-16 06:09:19,062 browser._LogBrowserInfo:132 disable_discard_framebuffer (INFO) 2016-08-16 06:09:19,062 browser._LogBrowserInfo:132 disable_dxgi_zero_copy_video (INFO) 2016-08-16 06:09:19,062 browser._LogBrowserInfo:132 disable_framebuffer_cmaa (INFO) 2016-08-16 06:09:19,062 browser._LogBrowserInfo:132 exit_on_context_lost (INFO) 2016-08-16 06:09:19,062 browser._LogBrowserInfo:132 scalarize_vec_and_mat_constructor_args
,
Aug 17 2016
I don't know whether or not they were using GPU before.
,
Aug 18 2016
I looked at the chart for memory:chrome:GPU_process:process_count_avg and there is no data reported after r405235. I also looked at the chart for memory:chrome:ALL_processes:process_count_avg and there's a sharp drop between r405236 and r405763, which I suspect is due to the fact that there's no GPU process anymore. I think it's safe to assume that this is due to a configuration change on the bots and we can mark it as WontFix. |
|||||||||||||||||||||
►
Sign in to add a comment |
|||||||||||||||||||||
Comment 1 by petrcermak@chromium.org
, Aug 8 2016