Ansible managed NTP config

[Kevin Stevens]

This is a feature request.

Sometimes our servers' times will drift apart enough to cause Neutron agents/Nova services to flap up and down. It would probably be a good idea to have the entire environment agree on time similar to the guidelines presented in the Openstack deployment guide.

Proposal: It would be nice if the Infra nodes had a configurable list of hosts to use as NTP servers, defaulting to the pool.ntp.org servers. By default, the infra nodes would be peers of each other. Any physical host *not* Infra would be clients of the Infra nodes.

[Miguel Grinberg (miguelgrinberg)]

I have suffered from this as well when trying to correlate log files for containers hosted on different infra nodes.

[Jesse Pretorius (jesse-pretorius)]

As this is a wishlist item, I'm removing the targeted milestone until someone picks it up.

[Major Hayden (rackerhacker)]

Just proposed a fix:

  https://review.openstack.org/#/c/225006/

I like the idea of having the compute node physical hosts get their time from one of the infrastructure hosts, but I haven't figured out a good way to do that without breaking all in one deployments. With AIO's, the physical host is an infrastructure host *and* a compute host. :/

[Major Hayden (rackerhacker)]

Email thread underway on the ML: http://lists.openstack.org/pipermail/openstack-dev/2015-September/074913.html

[OpenStack Infra (hudson-openstack)]

Change abandoned by Major Hayden (<email address hidden>) on branch: master
Review: https://review.openstack.org/225006
Reason: See previous comment.

[Major Hayden (rackerhacker)]

Abandoned that change based on ML thread. Will examine the NTP role from upstream Ansible Galaxy.

[Bjoern (bjoern-t)]

As a ops guy I support having NTP configuration support inside openstack-ansible.
Ideally I would like to configure NTP in a way that all openstack nodes synchronize to the controller and the controller synchronize to upstream. The reason for this is, if upstream gets out of sync or worse, we still have all openstack nodes from a cloud consistent to each other. That prevents time drifts, if all systems would go directly to upstream and that would make log reading quite difficult

[Major Hayden (rackerhacker)]

This was added in openstack-ansible-security.