Cause make start to wait for master start.

Cause make start to wait for master start.

Note that this doesn't fix the root problem -- lack of a mutex covering buildbot start to twistd.pid creation. Still, master manager has its own mutex covering that process. The potential issue is a human issuing a make start exactly when master manager is, which is hopefully unlikely enough to be an acceptable failure mode.

BUG=588118
Committed: http://src.chromium.org/viewvc/chrome?view=rev&revision=299373

[ghost stip (do not use), 2016-03-17 23:56:47]

Description was changed from

==========
Cause make start to wait for master start.

Note that this doesn't fix the root problem -- lack of a mutex covering buildbot
start to twistd.pid creation. Still, master manager has its own mutex covering
that process. The protential issue is a human issuing a make start exactly when
master manager is, which is hopefully unlikely enough to be an acceptable
failure mode.

BUG=588118
==========

to

==========
Cause make start to wait for master start.

Note that this doesn't fix the root problem -- lack of a mutex covering buildbot
start to twistd.pid creation. Still, master manager has its own mutex covering
that process. The protential issue is a human issuing a make start exactly when
master manager is, which is hopefully unlikely enough to be an acceptable
failure mode.

BUG=588118
==========

[ghost stip (do not use), 2016-03-17 23:56:47]

stip@chromium.org changed reviewers:
+ dsansome@chromium.org

[ghost stip (do not use), 2016-03-17 23:56:53]

ptal

[dsansome, 2016-03-18 00:01:57]

https://chromiumcodereview.appspot.com/1808383002/diff/1/masters/master-commo...
File masters/master-common-rules.mk (right):

https://chromiumcodereview.appspot.com/1808383002/diff/1/masters/master-commo...
masters/master-common-rules.mk:74: while `test ! -f twistd.pid`; do sleep 1;
done;
What if twistd.pid is never created?  Does master manager have its own timeout?

[ghost stip (do not use), 2016-03-18 00:05:39]

Yes. 10 minutes for each master_manager instance (one for each master), and an
11 minute overall timer for each cron-launched master_manager_launcher instance
(one for each bot, launches the individual master_manager instances):
https://goto.google.com/ctxuh

[ghost stip (do not use), 2016-03-18 00:06:09]

Description was changed from

==========
Cause make start to wait for master start.

Note that this doesn't fix the root problem -- lack of a mutex covering buildbot
start to twistd.pid creation. Still, master manager has its own mutex covering
that process. The protential issue is a human issuing a make start exactly when
master manager is, which is hopefully unlikely enough to be an acceptable
failure mode.

BUG=588118
==========

to

==========
Cause make start to wait for master start.

Note that this doesn't fix the root problem -- lack of a mutex covering buildbot
start to twistd.pid creation. Still, master manager has its own mutex covering
that process. The potential issue is a human issuing a make start exactly when
master manager is, which is hopefully unlikely enough to be an acceptable
failure mode.

BUG=588118
==========

[dsansome, 2016-03-18 00:10:13]

lgtm

[ghost stip (do not use), 2016-03-18 00:24:59]

The CQ bit was checked by stip@chromium.org

[commit-bot: I haz the power, 2016-03-18 00:25:04]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1808383002/1
View timeline at
 https://chromium-cq-status.appspot.com/patch-timeline/1808383002/1

[commit-bot: I haz the power, 2016-03-18 00:29:58]

Description was changed from

==========
Cause make start to wait for master start.

Note that this doesn't fix the root problem -- lack of a mutex covering buildbot
start to twistd.pid creation. Still, master manager has its own mutex covering
that process. The potential issue is a human issuing a make start exactly when
master manager is, which is hopefully unlikely enough to be an acceptable
failure mode.

BUG=588118
==========

to

==========
Cause make start to wait for master start.

Note that this doesn't fix the root problem -- lack of a mutex covering buildbot
start to twistd.pid creation. Still, master manager has its own mutex covering
that process. The potential issue is a human issuing a make start exactly when
master manager is, which is hopefully unlikely enough to be an acceptable
failure mode.

BUG=588118

Committed: http://src.chromium.org/viewvc/chrome?view=rev&revision=299373
==========

[commit-bot: I haz the power, 2016-03-18 00:29:59]

Committed patchset #1 (id:1) as
http://src.chromium.org/viewvc/chrome?view=rev&revision=299373