Issue metadata
Sign in to add a comment
|
battor.trivial_pages failing on Android Nexus6 WebView Perf |
||||||||||||||||||||||
Issue descriptionbattor.trivial_pages failing on 1 builders Builders failed on: - Android Nexus6 WebView Perf: https://build.chromium.org/p/chromium.perf/builders/Android%20Nexus6%20WebView%20Perf The baffling part about this, at least for the Nexus6 WebView Perf failures, is that the benchmark seems to be regularly failing with the following error: Traceback (most recent call last): File "../../third_party/catapult/devil/devil/android/tools/system_app.py", line 218, in <module> sys.exit(main(sys.argv[1:])) File "../../third_party/catapult/devil/devil/android/tools/system_app.py", line 208, in main devices = script_common.GetDevices(args.devices, args.blacklist_file) File "/b/swarming/w/ir/third_party/catapult/devil/devil/android/tools/script_common.py", line 47, in GetDevices raise device_errors.NoDevicesError() devil.android.device_errors.NoDevicesError: No devices attached. However, other benchmarks that use the same device (--device5) are able to run successfully, which sure doesn't make it *look* like a device offline error. I'm going to look into this.
,
Aug 9 2017
Ben, do you have any idea what might be happening here? I'm seeing just this one benchmark fail with a devil NoDevicesError, but other benchmarks that run on the same device (--device5) seem to be succeeding with no problem. This might make sense for a single run, but it seems to happen regularly. I have no idea why a device might regularly go offline for just a single benchmark. Here are a couple runs that exhibit this behavior: https://build.chromium.org/p/chromium.perf/builders/Android%20Nexus6%20WebView%20Perf/builds/110 https://build.chromium.org/p/chromium.perf/builders/Android%20Nexus6%20WebView%20Perf/builds/107
,
Aug 9 2017
,
Aug 9 2017
It looks like swarming is taking the device offline right before these tests. There's a whole lot of "Devices set([u'ZX1G22LHMW']) dropped offline during pre-task cleanup." in the logs. It also seems to be affecting v8.browsing_mobile on that bot as well. I landed https://chrome-internal-review.googlesource.com/c/427131 yesterday which might help with some of this. I see the bot has had two successful battor.trivial_pages runs since it landed. Maybe the change did improve things. Let's continue to watch it and see how it does. *crosses fingers*
,
Aug 10 2017
I'll add a NextAction date of Monday to check back and see if battor.trivial_pages is still having problems.
,
Aug 14 2017
The NextAction date has arrived: 2017-08-14
,
Aug 14 2017
This no longer appears to be happening on Android. Closing this bug. Thanks for your help Ben! |
|||||||||||||||||||||||
►
Sign in to add a comment |
|||||||||||||||||||||||
Comment 1 by charliea@chromium.org
, Aug 9 2017