net_unittests runtime increased by about 50% |
|||
Issue descriptionThis CL seems to be where the change happened. https://codereview.chromium.org/2889713002/ It seems that 252 SendAlgorithmTests (up from 42) are running. Many of these tests take ~400ms, so this suite now takes about 2 minutes to run. Prior to this CL, net_unittests took 4+ minutes on iOS. Now it takes 6+ minutes. These times were measured locally on a macbook. I also glanced at a Windows run before that CL landed and today, and it looks that there are now 252 tests, most of which take close to a half second. Run from Jun 2 (252 tests): https://luci-logdog.appspot.com/v/?s=chromium%2Fbb%2Fchromium.win%2FWin10_Tests_x64%2F12350%2F%2B%2Frecipes%2Fsteps%2Fnet_unittests_on_Windows-10-10586%2F0%2Fstdout Run from May 15 (42 tests): https://luci-logdog.appspot.com/v/?s=chromium%2Fbb%2Fchromium.win%2FWin10_Tests_x64%2F11680%2F%2B%2Frecipes%2Fsteps%2Fnet_unittests_on_Windows-10-10586%2F0%2Fstdout Are there any guidelines for increasing test execution time by this much? A 50% increase in a test that is run multiple times on the CQ is negatively affecting iOS capacity.
,
Jun 2 2017
Generally, "don't do it without warning us" is the guideline :).
,
Jun 2 2017
,
Jun 3 2017
The following revision refers to this bug: https://chromium.googlesource.com/chromium/src.git/+/05bfd593fef2c380d3d37181069d6212eb8cd5aa commit 05bfd593fef2c380d3d37181069d6212eb8cd5aa Author: rch <rch@chromium.org> Date: Sat Jun 03 00:52:11 2017 Speed up SendAlgorithmTest by reducing the number of cubic fix permutations that are run to be either "All" or "None" Merge internal change: 157890562 BUG= 729167 Review-Url: https://codereview.chromium.org/2920063002 Cr-Commit-Position: refs/heads/master@{#476844} [modify] https://crrev.com/05bfd593fef2c380d3d37181069d6212eb8cd5aa/net/quic/core/congestion_control/send_algorithm_test.cc
,
Jun 6 2017
rch: I assume this is fixed?
,
Jun 6 2017
Oh, indeed. Thanks.
,
Jun 6 2017
I checked and the runtime of net_unittests has returned to its previous level on iOS. I'm also monitoring capacity and there haven't been any spikes yet. Thanks for the quick fix! |
|||
►
Sign in to add a comment |
|||
Comment 1 by dpranke@chromium.org
, Jun 2 2017