Open Stack

-----Original Message-----
From: John Griffith <john.griffith8 at gmail.com>
Reply: OpenStack Development Mailing List (not for usage questions) <openstack-dev at lists.openstack.org>>
Date: September 25, 2014 at 12:27:52
To: OpenStack Development Mailing List (not for usage questions) <openstack-dev at lists.openstack.org>>
Subject:  Re: [openstack-dev] [Ironic] Get rid of the sample config file

> On Thu, Sep 25, 2014 at 12:34 PM, Devdatta Kulkarni <
> devdatta.kulkarni at rackspace.com> wrote:
>  
> > Hi,
> >
> > We have faced this situation in Solum several times. And in fact this was
> > one of the topics
> > that we discussed in our last irc meeting.
> >
> > We landed on separating the sample check from pep8 gate into a non-voting
> > gate.
> > One reason to keep the sample check is so that when say a feature in your
> > code fails
> > due to some upstream changes and for which you don't have coverage in your
> > functional tests then
> > a non-voting but failing sample check gate can be used as a starting point
> > of the failure investigation.
> >
> > More details about the discussion can be found here:
> >
> > http://eavesdrop.openstack.org/meetings/solum_team_meeting/2014/solum_team_meeting.2014-09-23-16.00.log.txt  
> >
> > - Devdatta
> >
> > ------------------------------
> > *From:* David Shrewsbury [shrewsbury.dave at gmail.com]
> > *Sent:* Thursday, September 25, 2014 12:42 PM
> > *To:* OpenStack Development Mailing List (not for usage questions)
> > *Subject:* Re: [openstack-dev] [Ironic] Get rid of the sample config file
> >
> > Hi!
> >
> > On Thu, Sep 25, 2014 at 12:23 PM, Lucas Alvares Gomes <
> > lucasagomes at gmail.com> wrote:
> >
> >> Hi,
> >>
> >> Today we have hit the problem of having an outdated sample
> >> configuration file again[1]. The problem of the sample generation is
> >> that it picks up configuration from other projects/libs
> >> (keystoneclient in that case) and this break the Ironic gate without
> >> us doing anything.
> >>
> >> So, what you guys think about removing the test that compares the
> >> configuration files and makes it no longer gate[2]?
> >>
> >> We already have a tox command to generate the sample configuration
> >> file[3], so folks that needs it can generate it locally.
> >>
> >> Does anyone disagree?
> >>
> >>
> > +1 to this, but I think we should document how to generate the sample
> > config
> > in our documentation (install guide?).
> >
> > -Dave
> > --
> > David Shrewsbury (Shrews)
> >
> > _______________________________________________
> > OpenStack-dev mailing list
> > OpenStack-dev at lists.openstack.org
> > http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
> >
> >
> I tried this in Cinder a while back and was actually rather surprised by
> the overwhelming push-back I received from the Operator community, and
> whether I agreed with all of it or not, the last thing I want to do is
> ignore the Operators that are actually standing up and maintaining what
> we're building.
>  
> Really at the end of the day this isn't really that big of a deal. It's
> relatively easy to update the config in most of the projects "tox
> -egenconfig" see my posting back in May [1]. For all the more often this
> should happen I'm not sure why we can't have enough contributors that are
> just pro-active enough to "fix it up" when they see it falls out of date.
>  
> John
>  
> [1]: http://lists.openstack.org/pipermail/openstack-dev/2014-May/036438.html  

+1 to what John just said.
 
I know in Keystone we update the sample config (usually) whenever we notice it out of date. Often we ask developers making config changes to run `tox -esample_config` and re-upload their patch. If someone misses we (the cores) will do a patch that just updates the sample config along the way. Ideally we should have a check job that just reports the config is out of date (instead of blocking the review).

The issue is the premise that there are 2 options:

1) Gate on the sample config being current
2) Have no sample config in the tree.

The missing third option is the proactive approach (plus having something convenient like `tox -egenconfig` or `tox -eupdate_sample_config` to make it convenient to update the sample config) is the approach that covers both sides nicely. The Operators/deployers have the sample config in tree, the developers don’t get patched rejected in the gate because the sample config doesn’t match new options in an external library.

I know a lot of operators and deployers appreciate the sample config being in-tree.

—Morgan