Openstack client puppet provider should retry failed commands

[Bogdan Dobrelya]

Openstack client (OSC) provider raises puppet error on a first keystone service error encountered, like 503 Service unavailable.
While the correct fix would be to ensure keystone service (haproxy backend) always availabe while issuing commands to OSC provider, this seems barely possible in Fuel deploy patterns. Hence, retries should present as well.

[Bogdan Dobrelya (bogdando)]

The deployment impact is critical and there are multiple duplicates.

[Bogdan Dobrelya (bogdando)]

Note, this bug must be first fixed for the puppet-openstacklib upstream

[Bogdan Dobrelya (bogdando)]

Note, the proper fix should be done here https://bugs.launchpad.net/fuel/+bug/1485644. Fixing that bug first should make this one obsolete as there wouldn't be unexpected backend downtimes.

[Ivan Berezovskiy (iberezovskiy)]

Upstream work should be done early at 8.0 timeline.

[Max Yatsenko (myatsenko)]

In related bug: https://bugs.launchpad.net/fuel/+bug/1485644 fix was successfully merged : https://review.openstack.org/#/c/214321/

[Bogdan Dobrelya (bogdando)]

Lowering to the medium for 8.0 as there were patches accepted which made retries not really needed

[Max Yatsenko (myatsenko)]

This patch adds retries to openstack command:

#Add retries to the openstack command
https://review.openstack.org/#/c/220224/

[Max Yatsenko (myatsenko)]

The patch: https://review.openstack.org/#/c/220224 was successfully merged.

[Ivan Berezovskiy (iberezovskiy)]

https://review.openstack.org/#/c/220224/ is merged

[Bogdan Dobrelya (bogdando)]

Great contribution example to benefit *all*. Kudos to all who participated!