Oh, I made a typo. We need n1-highmem-32. LTO eats tons of RAM.

+abw for approval

Note: This request represents a total of 960 cores

This is correct. See also https://crbug.com/598011, where bare metal bots were upgraded. This is a GCE counterpart.

LTO is a costful feature (at least now, while ThinLTO is still in its infancy), but these costs for this particular launch (LTO and then CFI on Linux) was approved by mal@, as the benefits seem worthwhile: faster Chrome (if built with LTO) and more secure Chrome when CFI is enabled (LTO is a prerequisite for CFI).

Acronyms used:

LTO: http://llvm.org/docs/LinkTimeOptimization.html
CFI: https://www.chromium.org/developers/testing/control-flow-integrity

friendly ping

Approved ! We love this project, it is worth every cent.

Thank you, Adam!

The following revision refers to this bug:
  https://chrome-internal.googlesource.com/infra/infra_internal.git/+/a560a24c8c35a38661b6875c367ed20ef96496ba

commit a560a24c8c35a38661b6875c367ed20ef96496ba
Author: johnw <johnw@google.com>
Date: Wed May 18 22:07:22 2016

Thank you, John. Adding these slaves to 'LTO Linux Perf':
https://codereview.chromium.org/1992083002/

I will give them to run on Chromium.FYI for a day and if we see it can keep up with the incoming CL traffic, I will move them to Chromium.Perf:
https://build.chromium.org/p/chromium.perf/builders/Linux%20Builder

This will actually 'free' 5 existing slaves.

The following revision refers to this bug:
  https://chrome-internal.googlesource.com/infra/infra_internal.git/+/8f0d6317db158033b21eadc640e896cc3bea7728

commit 8f0d6317db158033b21eadc640e896cc3bea7728
Author: johnw <johnw@google.com>
Date: Wed May 18 23:34:33 2016

The following revision refers to this bug:
  https://chromium.googlesource.com/chromium/tools/build.git/+/4b00862ef7c87ecd73c5e40de4c8a1be097b8dbf

commit 4b00862ef7c87ecd73c5e40de4c8a1be097b8dbf
Author: krasin@google.com <krasin@google.com>
Date: Wed May 18 23:53:05 2016

Add 30 more slaves into 'LTO Linux Perf' buildbot.

The goal is to verify that we can (on average) have a build for every Chromium CL.
If that's true, these slaves will be moved to the real Perf Linux Builder:
https://build.chromium.org/p/chromium.perf/builders/Linux%20Builder

and then LTO will be launched on Linux.

BUG= 453195 , 612214 

Review-Url: https://codereview.chromium.org/1992083002

git-svn-id: svn://svn.chromium.org/chrome/trunk/tools/build@300641 0039d316-1c4b-4281-b951-d872f2087c98

[modify] https://crrev.com/4b00862ef7c87ecd73c5e40de4c8a1be097b8dbf/masters/master.chromium.fyi/slaves.cfg
[modify] https://crrev.com/4b00862ef7c87ecd73c5e40de4c8a1be097b8dbf/scripts/slave/compile.py

Slaves are granted. After chromium.fyi master restart (hopefully, EOD today), they will start building for 'LTO Linux Perf' bot. If everything goes well, I will switch them to the real Linux Perf builders tomorrow. That gives a chance to launch LTO on Friday, which likely means Monday to avoid unnecessary surprises.

The following revision refers to this bug:
  http://goto.ext.google.com/viewvc/chrome-internal?view=rev&revision=88029

------------------------------------------------------------------
r88029 | recipe-roller@chromium.org | 2016-05-19T00:03:10.287391Z

-----------------------------------------------------------------