New issue
Advanced search Search tips

Issue 903259 link

Starred by 1 user

Issue metadata

Status: Fixed
Owner:
Closed: Nov 8
Cc:
EstimatedDays: ----
NextAction: ----
OS: ----
Pri: 2
Type: Bug



Sign in to add a comment

mac_chromium_asan_rel_ng uses goma but builds with -j40

Project Member Reported by h...@chromium.org, Nov 8

Issue description

Labels: -OS-Mac
https://ci.chromium.org/p/chromium/builders/luci.chromium.try/linux_chromium_msan_rel_ng/1286 builds with -j80

https://ci.chromium.org/p/chromium/builders/luci.chromium.try/linux_chromium_chromeos_msan_rel_ng/1044 builds with -j150

https://ci.chromium.org/p/chromium/builders/luci.chromium.try/win-asan/104 builds with -j80


They all take at least 1 hour for a full build. Is there a good reason these aren't closer to -j1000?
Cc: ukai@chromium.org tikuta@chromium.org
If I understand correctly, the -j option is decided by this file:

https://cs.chromium.org/chromium/src/infra/config/global/cr-buildbucket.cfg?rcl=ae58e0551120acd2d198d91397f83c43a3e214fa&l=95

And some bots there are configured for higher -j numbers via the 


+tikuta and ukai, who have landed patches to bump these numbers before (crrev.com/605274 crrev.com/579317).

Can we just increase the numbers, or do we need stronger machines?
Status: Started (was: Untriaged)
Patch: https://chromium-review.googlesource.com/c/chromium/src/+/1326153
Our main concern is backend capacity.
Also goma client itself consumes some amount of local resource in proportion to parallelism. And most bots do not have large resources like developer's beefy machine.

So I think there is scale limit less than -j1000 in CQ machine.
Project Member

Comment 5 by bugdroid1@chromium.org, Nov 8

The following revision refers to this bug:
  https://chromium.googlesource.com/chromium/src.git/+/056e148a4d91c17275aa8c9216566c463014c306

commit 056e148a4d91c17275aa8c9216566c463014c306
Author: Hans Wennborg <hans@chromium.org>
Date: Thu Nov 08 13:10:51 2018

run ninja with higher -j for bots blocking clang rolls

These bots often time out during clobber builds, blocking clang rolls.
Bump the -j number to those used by similar bots.

Bug:  903259 
Change-Id: I4ea85b111faf9bf0a441c18730a6a8d0bee20d23
Reviewed-on: https://chromium-review.googlesource.com/c/1326153
Commit-Queue: Hans Wennborg <hans@chromium.org>
Reviewed-by: Takuto Ikuta <tikuta@chromium.org>
Cr-Commit-Position: refs/heads/master@{#606437}
[modify] https://crrev.com/056e148a4d91c17275aa8c9216566c463014c306/infra/config/global/cr-buildbucket.cfg

Owner: h...@chromium.org
Trying a whitespace clobber change here to see the effect: https://chromium-review.googlesource.com/c/chromium/src/+/1325991
This is still using -j40 :-(
https://ci.chromium.org/p/chromium/builders/luci.chromium.try/mac_chromium_asan_rel_ng/5134

Do I need to do anything else to make it take effect?
Hey, the next build uses -j150
https://ci.chromium.org/p/chromium/builders/luci.chromium.try/linux_chromium_msan_rel_ng/1293
Maybe it just takes a while to propagate or something.
Status: Fixed (was: Started)
https://ci.chromium.org/p/chromium/builders/luci.chromium.try/mac_chromium_asan_rel_ng/5135 is running with -j150

Let's call this done.

Sign in to add a comment