notifier_kombu should not retry forever

[Erica Windisch]

There is a rabbit_max_retries option, but this is not working correctly.

First, the configvar is defined as a StrOpt. This means that the check against the integer retry_attempts will fail.

Secondly, rabbit_max_retries is tested as a boolean before comparing against retry_attempts. With the value 0, this would be false and the setting rabbit_max_retries=0 would actually retry indefinitely, which is not the obvious behavior.

FYI - It turns out that this code is blocking. If notifications cannot be sent, if a RabbitMQ server cannot be reached, Glance will simply block on attempting to send notifications, while API requests timeout. As such, until that is rectified, having the max_retries functional is quite handy.

[OpenStack Infra (hudson-openstack)]

Fix proposed to branch: master
Review: https://review.openstack.org/23980

[OpenStack Infra (hudson-openstack)]

Reviewed: https://review.openstack.org/23980
Committed: http://github.com/openstack/glance/commit/1e607c9c3f34345af4bc18b0822606015ba36596
Submitter: Jenkins
Branch: master

commit 1e607c9c3f34345af4bc18b0822606015ba36596
Author: Eric Windisch <email address hidden>
Date: Fri Mar 8 23:46:36 2013 -0500

    Fix rabbit_max_retry

    First, the configvar is defined as a StrOpt. This means that the check
    against the integer retry_attempts will fail.

    Secondly, rabbit_max_retries is tested as a boolean before comparing
    against retry_attempts. With the value 0, this would be false and the
    setting rabbit_max_retries=0 would actually retry indefinitely, which is
    not the obvious behavior. This patch will make the default not to
    retry at all (which was the implied, but not actual, behavior)

    Fixes bug 1152866

    Change-Id: Ic39d708f0294fc5d919cb2a5d9fae532128ef17e