Add 10% experiment for mac_chromium_10.10_rel_ng

Add 10% experiment for mac_chromium_10.10_rel_ng

This builder is green and will be added to the CQ soon.

BUG=498330
Committed: https://crrev.com/995fe62f5bfcd8f541187c44c01cf7ec2d63c92e
Cr-Commit-Position: refs/heads/master@{#333526}

[smut, 2015-06-09 16:53:34]

smut@google.com changed reviewers:
+ phajdan.jr@chromium.org, sergeyberezin@chromium.org, sergiyb@chromium.org,
sheyang@chromium.org

[Sergey Berezin, 2015-06-09 17:03:10]

This is not going to fly - you only have 2 slaves in the pool. Judging by the
existing builds cycle time, you need at least 20.

I'd suggest trying 10% experiment, while upgrading the pool.

[smut, 2015-06-09 17:11:22]

On 2015/06/09 17:03:10, Sergey Berezin wrote:
> This is not going to fly - you only have 2 slaves in the pool. Judging by the
> existing builds cycle time, you need at least 20.
> 
> I'd suggest trying 10% experiment, while upgrading the pool.

Is there any harm in experimental jobs being backed up? Why not set it to 100%,
then scale the pool until there's no backlog?

[Sergey Berezin, 2015-06-09 17:17:23]

On 2015/06/09 17:11:22, smut wrote:
> On 2015/06/09 17:03:10, Sergey Berezin wrote:
> > This is not going to fly - you only have 2 slaves in the pool. Judging by
the
> > existing builds cycle time, you need at least 20.
> > 
> > I'd suggest trying 10% experiment, while upgrading the pool.
> 
> Is there any harm in experimental jobs being backed up? Why not set it to
100%,
> then scale the pool until there's no backlog?

I'd rather not do that. The jobs will pile up unbelievably fast, and you'd have
to clean up 1000s of stale jobs later anyway, even with a bigger pool. It may
also be taxing on the buildbot to display the builder's page with 1000s of
pending builds (it happened before - the page wouldn't even load).

Having said that, it *might* be better with buildbucket now, since stale pending
jobs should just expire, but you'll likely still have a metric crapton of them.
Check with nodir@.

[smut, 2015-06-09 17:19:02]

On 2015/06/09 17:17:23, Sergey Berezin wrote:
> On 2015/06/09 17:11:22, smut wrote:
> > On 2015/06/09 17:03:10, Sergey Berezin wrote:
> > > This is not going to fly - you only have 2 slaves in the pool. Judging by
> the
> > > existing builds cycle time, you need at least 20.
> > > 
> > > I'd suggest trying 10% experiment, while upgrading the pool.
> > 
> > Is there any harm in experimental jobs being backed up? Why not set it to
> 100%,
> > then scale the pool until there's no backlog?
> 
> I'd rather not do that. The jobs will pile up unbelievably fast, and you'd
have
> to clean up 1000s of stale jobs later anyway, even with a bigger pool. It may
> also be taxing on the buildbot to display the builder's page with 1000s of
> pending builds (it happened before - the page wouldn't even load).
> 
> Having said that, it *might* be better with buildbucket now, since stale
pending
> jobs should just expire, but you'll likely still have a metric crapton of
them.
> Check with nodir@.

Changed to 10%.

[Sergiy Byelozyorov, 2015-06-09 17:38:21]

lgtm

[Sergey Berezin, 2015-06-09 17:41:04]

LGTM, thanks!

[smut, 2015-06-09 17:41:49]

The CQ bit was checked by smut@google.com

[commit-bot: I haz the power, 2015-06-09 17:42:58]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1171063002/20001

[commit-bot: I haz the power, 2015-06-09 18:06:46]

Committed patchset #2 (id:20001)

[commit-bot: I haz the power, 2015-06-09 18:07:45]

Patchset 2 (id:??) landed as
https://crrev.com/995fe62f5bfcd8f541187c44c01cf7ec2d63c92e
Cr-Commit-Position: refs/heads/master@{#333526}