[mongodb] Message posts are not always atomic

[Kurt Griffiths]

Currently, if we lose our network connection to mongod we will return a 503 to the client and a partial enqueuing of the posted message batch is possible.

[OpenStack Infra (hudson-openstack)]

Reviewed: https://review.openstack.org/67320
Committed: https://git.openstack.org/cgit/openstack/marconi/commit/?id=fc71f6a17e27155b55bd80aa8b67881ed8b6a7a9
Submitter: Jenkins
Branch: master

commit fc71f6a17e27155b55bd80aa8b67881ed8b6a7a9
Author: kgriffs <email address hidden>
Date: Thu Jan 16 17:01:28 2014 -0600

    fix(mongodb): Ensure batch message posts are atomic (all or nothing)

    In the case of a network failure it is not possible to discover which,
    if any, of the messages in a given batch were inserted before the
    outage occurred.

    This patch works around the problem by inserting messages in two
    phases:

    1. Assign all messages a "batch" or "group" ID and insert them
    2. Once it has been confirmed that *all* messages were inserted
       correctly, finalize the transaction by clearing the batch
       ID, 'g', from the messages just inserted.

    Message listing and counting were also modified to ignore any
    non-finalized messages, i.e., those messages that have a non-null
    'g' value.

    Orphaned messages will eventually expire and be cleaned up as
    usual via the TTL index.

    Closes-Bug: #1270263
    Change-Id: I9de61cdfdf6dfcbe07ef76bf63a293091f087663