[02:55] <larsks> rharper: that looks better, yes.
[03:05] <larsks> rharper: although I note that I get KeyError: 'r' when trying to run 'make srpm'...
[03:08] <larsks> ...because I needed PYVER=python2, right.
[03:09] <larsks> Although: I would call it a bug if something declares "template: cheetah" and we continue without producing an error message...
=== sambetts|afk is now known as sambetts
[14:46] <rharper> larsks: thanks for the feedback, we're looking to drop cheetah as a template (there is some code which does fallback if you don't have cheetah but those messages go to logging, which isn't configured for the tool)
[14:46] <rharper> so you're right w.r.t the error; and we're looking to drop cheetah soon
[17:50] <blackboxsw> hiya larsks, do you have any idea what meta package I'd need in centos/7 to setup repositories which would give me access to download python3 system package for oauthlib? I've got epel-release installed, but it doesn't yum search is coming up empty.
[17:51] <blackboxsw> sorry larsks, I should introduce myself. I'm Chad Smith, a recent join to Canonical's cloud-init/curtin team.
[17:52] <larsks> blackboxsw: hi there!
[17:53] <larsks> I know that oauthlib is not available in RHEL.  It may not be in CentOS either.  Let me look.
[17:54] <larsks> blackboxsw: yeah, there is no generally available oauthlib package.
[17:55] <blackboxsw> ok thanks for the confirmation larsks.
[17:55] <larsks> It is available in the openstack repositories (e.g., centos-release-openstack-newton), but you wouldn't normally expect those to be enabled where people are running cloud-init.
[17:56] <rharper> http://oauthlib.readthedocs.io/en/latest/installation.html  says it has a python3 oauthlib
[17:56] <rharper> for redhat
[17:56] <rharper> but that doesn't seam right, I would expected python34 or something
[18:10] <larsks> rharper: we actually include a patch in our cloud-init package specfically to fail gracefully if oauthlib is missing.
[18:10] <rharper> ah, ok
[18:10] <larsks> (we catch the importerror, and then raise an exception later on if something actually tires to call the oauth_headers method)
[18:10] <rharper> we probably want that then
[18:11] <larsks> http://chunk.io/f/c756394a7c91432f8a92fdbad2682fda
[18:12] <rharper> cool
[18:13] <blackboxsw> thanks larsks that works for me. I was just working on setting up a make ci-deps-centos target and I didn't want to rule out installing any dependencies I could get together if they are reasonably available for setting up our test environments.
[18:14] <blackboxsw> so not having python3-oathlib isn't a deal breaker for our CI
[18:14] <larsks> fwiw, it's a good idea not to depend on anything not available in the stock centos repositories.  E.g., don't rely on EPEL.
[18:18] <blackboxsw> +1
[18:52] <rharper> https://bugzilla.redhat.com/show_bug.cgi?id=1417025
[18:52] <ubot5> bugzilla.redhat.com bug 1417025 in cloud-init "cloud-init tries to run hostnamectl before dbus is up" [Unspecified,Closed: rawhide]
[18:52] <rharper> that's going to be "fun"
[19:57] <larsks> rharper: looks like that was fixed.
[20:10] <rharper> larsks: ok;  I was looking at ubuntu and our dbus.service doesn't depend on After=sysinit.target ;  was going to test with that
[20:11] <larsks> rharper: I finally threw up my hands and completely abandoned the systemd units in the source tree.
[20:11] <rharper> =/  ok, I can look at the changes;  we may need to template the units as well
[20:11] <larsks> There were too many weird edge cases where ubuntu and redhat had different dependencies.
[20:11] <rharper> yeah
[20:12] <larsks> Well, I had another suggestion:
[20:12] <larsks> Using systemd drop-ins for the distribution specific stuff.
[20:12] <rharper> ah, yes
[20:12] <rharper> that is a good idea
[20:12] <rharper> I need to play with those some more
[20:12] <larsks> THe model would be: cloud-init units only have dependencies on other cloud-init units, and dependencies on anything else would be configured in drop-ins.
[20:19] <rharper> yes, that makes sense