The current cron job for purge is executed at the precisely same time,
if we deploy multiple controller nodes with database clusterd,
for example using Galera cluster, it could cause DB spike.
This patch introduces a randomized sleep before executing purge job,
to avoid such kind of collision.
This commit is a kind of cherry-pick of the commit in puppet-nova
with change id I093a713a4f0ba48686de615aa8cb22b17a56917b .
Co-Authored-By: Rajesh Tailor <ratailor@redhat.com>
Change-Id: I36cf5f1e66fed580786970d9a953d983539a6e43
(cherry picked from commit I36cf5f1e66fed580786970d9a953d983539a6e43)
79ed64e355