"Out of capacity" on Windows perf bots when Microsoft pushes an update |
|||||||
Issue descriptionLink to buildbot status page: https://ci.chromium.org/buildbot/chromium.perf/Win%2010%20Perf/ Looks like the swarming bot may be down?
,
Mar 7 2018
I'm doing this for crbug.com/819780 so I'll just include this one.
,
Mar 7 2018
,
Mar 8 2018
The following revision refers to this bug: https://chromium.googlesource.com/chromium/src.git/+/56fd063683018627c56f813874e1ba28661e8baa commit 56fd063683018627c56f813874e1ba28661e8baa Author: Simon <simonhatch@chromium.org> Date: Thu Mar 08 18:14:46 2018 Change dimensions for win high-DPI They seem to have changed on the bot, changing here to match. Bug: 819780 , 818877 Change-Id: I305892716bcb337df71a1f1e2d0937c0db8541b6 Reviewed-on: https://chromium-review.googlesource.com/953379 Reviewed-by: Emily Hanley <eyaich@chromium.org> Commit-Queue: Simon Hatch <simonhatch@chromium.org> Cr-Commit-Position: refs/heads/master@{#541839} [modify] https://crrev.com/56fd063683018627c56f813874e1ba28661e8baa/testing/buildbot/chromium.perf.json [modify] https://crrev.com/56fd063683018627c56f813874e1ba28661e8baa/tools/perf/core/perf_data_generator.py
,
Mar 9 2018
,
Mar 14 2018
Reopening this because we still need to do this for Win 10 Perf. I'll go ahead and do that.
,
Mar 14 2018
+nednguyen@ FYI I talked some to bevc@ about this. It looks like the problem is that sometimes, even if Windows Update is disabled, Microsoft pushes OS updates, which change the version number. Because we use that version number as one of the criteria in our swarming target, our benchmarks can no longer find a machine to run on and we start getting out of capacity errors. You can see this clearly on build117-b1: it was at version Windows-10-16299.248 12 hours or so ago (http://bit.ly/2FKUmLJ, search for Windows-10-16299.248) but now seems to be at Windows-10-16299.309 (http://bit.ly/2DsRcud, search for Windows-10-16299.309) and experiencing the same "out of capacity" errors that the other bots were previously experiencing. Ned suggested that we might not want to target based on the OS version at all (it's redundant - we already specify the bots that we want to target). However, we need to make sure that not specifying the target OS version doesn't have some confusing results. Apparently if you omit the GPU targeting criteria, for example, it means that you need to run on a machine with *no* GPU, rather than being able to run on a machine with *any* GPU.
,
Mar 14 2018
Stupid question, since you specify the device id's in perf_data_generator, why do you need any of those other dimensions?
,
Mar 14 2018
#8: we don't need most of other dimensions, except for gpu dimension (see 36 --> 40 in https://bugs.chromium.org/p/chromium/issues/detail?id=717744#c36)
,
Mar 14 2018
,
Mar 15 2018
The following revision refers to this bug: https://chromium.googlesource.com/chromium/src.git/+/7650de1d462a76c0e41d924177caa823fc78cebf commit 7650de1d462a76c0e41d924177caa823fc78cebf Author: Charlie Andrews <charliea@chromium.org> Date: Thu Mar 15 17:33:35 2018 Update the Windows target OS version strings to reflect system update Windows silently updated and, because we target the exact version, the benchmarks running on these machines started to fail with "out of capacity" errors. The long term solution here probably has to do with removing the OS version from the targeting criteria altogether given the fact that we already target the exact machines, but in the short term, this should get our bots greener. TBR=nednguyen@google.com Bug: 818877 Change-Id: I95ecfc8b3381eaa45c0d69d2b0caa747d1b1f9eb Reviewed-on: https://chromium-review.googlesource.com/963146 Commit-Queue: Ned Nguyen <nednguyen@google.com> Reviewed-by: Ned Nguyen <nednguyen@google.com> Reviewed-by: Charlie Andrews <charliea@chromium.org> Cr-Commit-Position: refs/heads/master@{#543427} [modify] https://crrev.com/7650de1d462a76c0e41d924177caa823fc78cebf/testing/buildbot/chromium.perf.json [modify] https://crrev.com/7650de1d462a76c0e41d924177caa823fc78cebf/tools/perf/core/perf_data_generator.py
,
Mar 16 2018
,
Mar 16 2018
Nice work, Charlie! |
|||||||
►
Sign in to add a comment |
|||||||
Comment 1 by martiniss@chromium.org
, Mar 6 2018Status: Assigned (was: Untriaged)