On Jul 10, 2013, at 5:01 AM, Joe Gordon <[email protected]> wrote:
> > > > On Tue, Jul 2, 2013 at 8:43 PM, Mark McLoughlin <[email protected]> wrote: > On Tue, 2013-07-02 at 17:48 +0000, Jeremy Stanley wrote: > > On 2013-07-01 15:10:26 -0700 (-0700), Mark Washenberger wrote: > > [...] > > > The talk about permanence confuses me, unless we mean that more > > > permanent values are overridden by less permanent ones. > > [...] > > > > I think the "permanence" counter argument (which I don't agree with, > > just recounting it for completeness) was that command-line arguments > > may be embedded in init scripts by some distributions and then > > administrators would be surprised when their modifications to the > > configuration files weren't respected. > > Yes, that was what the "permanence" discussion related to. The example I > was thinking of was '--logfile /var/log/nova/api.log' which doesn't seem > like a ridiculous thing to pass via the command line. > > Since we've clearly moved on, I'm not sure replaying old points is very > constructive, but you have hit on an interesting topic, so ... :) > > > Ultimately, however, any time > > distribution defaults which could be set in packaged configuration > > are instead being set with the service command-line in packaged init > > scripts, I would tend to just consider that a (serious) packaging > > bug and certainly nothing we should be catering to as a project. > > That's very ... stringent. But I do mostly agree. Distros shouldn't > stick a tonne of distro defaults on the command line of services. > > The two principles that matter IMHO are: > > 1) users should be able to override defaults > > 2) if a user deletes their config file, they get back to the defaults > > Something we've experimented with in Red Hat OpenStack is to put distro > defaults in e.g. /usr/share/nova/nova-dist.conf. You can see some of the > thinking here, for example: > > https://bugzilla.redhat.com/show_bug.cgi?id=887334#c4 > > The alternative approach would be to patch the code with distro > defaults. > > Now, you could say that distros shouldn't need to modify the defaults. > That's fair, but I don't think there's anything too crazy in our distro > defaults either: > > https://github.com/redhat-openstack/openstack-nova/blob/master/nova.conf > > > I took a look at those and most of the values you have are already the > current defaults. The only one that stands out is: > > > force_dhcp_release = True > > So my question here, is do you think 'force_dhcp_release = False' is a sane > default or should the default be True like you have it? I would like to see this as True for sure. Historically it was was false because it requires a special binary from dnsmasq that wasn't packaged by the distros. Vish > > > > > Anyone got other thoughts on how distros should handle this? > > Cheers, > Mark. > > > _______________________________________________ > OpenStack-dev mailing list > [email protected] > http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev > > _______________________________________________ > OpenStack-dev mailing list > [email protected] > http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
_______________________________________________ OpenStack-dev mailing list [email protected] http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
