This started as an effort in the all-hands hackathon. Let's continue working toward ansible scripts for setting controller and compute node configuration. Would be nice be able to choose based on distro and version. Current list would be:
RDO kilo
RDO liberty
Fuel 8
The hackathon code (attached) was focused on RDO kilo, and should be a good starting point for the others.
we no longer need the pipeline.yml file for ceilometer. it was nice to have in kilo, but our UDP output breaks liberty
Removed the template. Let me know what options in, I assume, ceilometer's api_paste.ini need to not be there.
we don’t do anything with trove, so we should remove any customizations for that
I had only left in the restart code, removed that.
there are other edits to rsyslog.conf to increase the message size and use a higher resolution date format for forwarded messages. we were replacing the file, but would be nice to edit them in to avoid losing any local customizations
It doesn't appear rsyslog cares if those options are in the main config or an included file, so I just added them to the templated file. Only would be a problem if the main file had conflicting options which the stock one doesn't appear to.
Questions:
i didn’t see where the various conf file settings are handled (nova.conf, cinder.conf, etc.). Is that still WIP, or did I miss it?
Same as the off-site just moved to scripts/config_openstack.sh I think this was in an external script since the internal ini provider for ansible doesn't support adding a single value to a list. There may be some 3rd party module to pull in to make this more generic so it can directly edit ini files and not need 'openstack-config'
I didn’t see any differentiation between kilo and liberty for config files. WIP? we would typically start with a vanilla config file from a pristine install and use that as the basis for the update. If they are verified to be the same, then no action required. If not, we should have templates for each stack version.
Currently it's in config_openstack.sh to make those decisions, obviously that will need to move to ansible code when we need to smart decisions in the playbook itself.
Would the intent be to initiate ansible from goldstone server, and have it apply change to the openstack hosts? If not, how should we package the ansible code to make it available for openstack hosts?
If so, what prerequisites do we have for goldstone-server and/or openstack hosts?
Running it from the goldstone server would be easiest, but really the only pre-requisite is ansible, I used 1.9.4 but I expect pretty much any version should work. This is really a simple playbook so all a user would need to do is copy the 'ansible' directory to something with ssh access to the openstack hosts and ansible installed.
This started as an effort in the all-hands hackathon. Let's continue working toward ansible scripts for setting controller and compute node configuration. Would be nice be able to choose based on distro and version. Current list would be:
The hackathon code (attached) was focused on RDO kilo, and should be a good starting point for the others.
hackathon-2016-Q1.zip