Dan Bode <[email protected]> writes:
> On Tue, May 11, 2010 at 3:27 PM, Nigel Kersten <[email protected]> wrote:
>> On Tue, May 11, 2010 at 3:19 PM, R.I.Pienaar <[email protected]> wrote: > >
>> ----- "Brian Gallew" <[email protected]> wrote: > >> in nodes and have them
>> propagate to their children, then I submit that >> nodes are so
>> fundamentally broken as to make node inheritance >> completely useless, in
>> which case node inheritance should be pulled. > > that sums it up well.
>>
>> ++
I was shocked how useless node inheritance was, once we actually hit the point
of doing more than adding a few static classes in it. :/
>> I haven't even told my users that nodes can do inheritance, as I haven't
>> yet come across any reason to do so.
We have two inherited definitions: default, and standard.
default applies to anything that connects to our network, and apples the
minimum set of classes required for conformance to our certification
standards, etc.
standard extends that with the behaviour we expect for most nodes: common
authentication setup, some additional monitoring services, and so forth.
Then, all nodes inherit one of those.
That way we don't have to repeat ourselves: to make something common to a
large class of nodes we derive their definition from default, and we can
change those.
So, we will be replacing that with a define, or parameterized class when they
ship, that allows us to do the same thing.
Given that node inheritance *looks* designed to achieve that goal, though,
and I can't imagine any other use for it, pulling the entire pointless feature
seems like the right decision to me.
> you guys don't use an external node classifier? Is there a reason you prefer
> to declare nodes?
If we have to hand-maintain the data then we may as well keep it inside the
puppet repository and language. There is *zero* benefit to fragmenting that
data when there is one, and only one, consumer of it.
We will likely move to an LDAP external node classifier once that actually
brings some benefit: when we have other services sharing that directory,
and/or a GUI interface to maintain it.
Until then, though, declaring nodes by hand is cheaper and quicker.
Daniel
For a small shop, or someone just starting out, I would encourage the same
thing, too. Until you grow to the size an external node classifier is
required there isn't much point spending the money to implement one — only to
discover you have to deploy version two that meets your actually needs.
--
✣ Daniel Pittman ✉ [email protected] ☎ +61 401 155 707
♽ made with 100 percent post-consumer electrons
--
You received this message because you are subscribed to the Google Groups
"Puppet Users" group.
To post to this group, send email to [email protected].
To unsubscribe from this group, send email to
[email protected].
For more options, visit this group at
http://groups.google.com/group/puppet-users?hl=en.
