Race condition causes resize-flavor to fail in Percona

[Steve Leon]

In Percona, when resizing an instance, Trove reports success (ACTIVE) even though percona hasnt truly finished up loading. ALthough mysql service is started and port 3306 is open, percona still does some tasks under the hood.

Normally this wouldnt matter that much. However when an instance is resized immediately after it has finished provisioning, the migrate fails and mysql gets corrupted.

To fix this issue, we need to make sure that percona mysql has shutdown successfully before doing a resize.

Since this is a race condition that happens when a resize is perform immediately after a instance-create finishes, this is only reproducible in automated tests.

[OpenStack Infra (hudson-openstack)]

Fix proposed to branch: master
Review: https://review.openstack.org/76446

[OpenStack Infra (hudson-openstack)]

Reviewed: https://review.openstack.org/76446
Committed: https://git.openstack.org/cgit/openstack/trove/commit/?id=c88c88fa3ddc2f171b6fec5cce3c09e268294b02
Submitter: Jenkins
Branch: master

commit c88c88fa3ddc2f171b6fec5cce3c09e268294b02
Author: Steve Leon <email address hidden>
Date: Tue Feb 25 22:06:58 2014 -0800

    Fixes a race condition in resize flavor for Percona

    When a resize happens immediately after an instances gets created,
    the mysql process doesnt have enough time to finished up
    provisioning before the resize-action invokes a stop and start
    resizing. Although the trove tries to stop mysql, the stop command
    doesnt ensure that mysql was stopped successfully. This causes
    mysql to be corrupted when migrated to a different VM. Normally
    this is not an issue since mysql provisioning has plenty of time
    to finish. This becomes an issue when doing an resize when mysql
    has just finished provisioned. Only happens in automated tests.
    This fix will ensure that the mysql server is stopped before a
    resize is performed.

    Change-Id: I0a3c351b61b82b88dfe05325a38ce93d6973a9f6
    Closes-Bug: 1284989