Autoscaling nested stack update incorrectly implemented

[Steven Hardy]

In a couple of places we've got this happening:

    updater = self.update_with_template(new_template,
                                                self._environment())
    updater.run_to_completion()
    self.check_update_complete(updater)

This means all update operations involving InstanceGroup and AutoScalingGroup resources will be serialized, since we're not following the correct pattern where we start the update in handle_update and poll for completion in check_create_complete.

This will become a more problematic issue when we move to decoupled nested stacks, since it means the parent stack will block on the nested stack update, when it could be doing something else and just periodically polling the nested stack state in check_update_complete instead

[Steven Hardy (shardy)]

Additional notes:

- stack-update initiated updates of InstanceGroup resource counts will block in the parent stack

- check_update_complete is only called once, so we don't actually wait for the update to complete when doing rolling updates etc.

[Zane Bitter (zaneb)]

Yep, this was originally only meant to happen when scaling was triggered from an alarm. Now that we're doing it during a rolling update, that's definitely bad for anything supposed to be happening in parallel.

[Steven Hardy (shardy)]

Oh wow, fixing this looks pretty hard :(

[Steven Hardy (shardy)]

Since fixing this looks hard/risky I'm going to defer to Kilo

[OpenStack Infra (hudson-openstack)]

Fix proposed to branch: master
Review: https://review.openstack.org/160595

[OpenStack Infra (hudson-openstack)]

Reviewed: https://review.openstack.org/160595
Committed: https://git.openstack.org/cgit/openstack/heat/commit/?id=f3e7fd5de217201777b70de182ab85c4542c7fc6
Submitter: Jenkins
Branch: master

commit f3e7fd5de217201777b70de182ab85c4542c7fc6
Author: Angus Salkeld <email address hidden>
Date: Thu Mar 5 10:49:50 2015 +1000

    Correct the check_update_completion logic in instance_group

    The automated resource update logic is incompatible with the up coming
    logic in decouple-nested, as it doesn't poll the check_update_complete
    correctly.

    Change-Id: I9c3d1f314ce7da3abda391ff2e0ab05d6c1d76ae
    Closes-bug: #1360292

[OpenStack Infra (hudson-openstack)]

Fix proposed to branch: master
Review: https://review.openstack.org/163743