hey folks! just wanted to pass along a note about this change I made to the infra scripts today:
https://infrastructure.fedoraproject.org/cgit/ansible.git/commit/?id=6c390c6...
What I did was extend a bit of scaffolding that already exists to handle fedmsg deployment differently if the host is in a group with 'python34-fedmsg' in its name (the 'python34' there is a EL-ism, but I didn't want to change it) so it now also covers fedmsg-hub as well as fedmsg-base.
Now, if a system is in a group with 'python34-fedmsg' in its name and the 'fedmsg/hub' role is run on that system, instead of enabling the 'fedmsg-hub' service, it'll enable the 'fedmsg-hub-3' service. The 'restart fedmsg-hub' handler is also tweaked so it conditionally restarts *both* 'fedmsg-hub.service' *and* 'fedmsg-hub-3.service' (this seemed like the easiest solution to the service restart problem; as we're using conditional-restart.sh, whichever isn't present will just be ignored).
This should help with moving things from Python 2 fedmsg to Python 3 fedmsg. It should be relatively easy to extend to the other fedmsg roles too, if desired - I didn't do that as I didn't want to change too much, and fedmsg-hub is all I needed for the box I'm migrating.
Please let me know if you have any comments, concerns or suggestions about this, thanks! I did ask around on IRC and it didn't seem like anyone had a better idea or plan for doing this, so I just went ahead...
infrastructure@lists.fedoraproject.org