[OpenStack-Infra] The future of our server naming patterns

James E. Blair corvus at inaugust.com
Wed May 25 16:31:01 UTC 2016


Spencer Krum <nibz at spencerkrum.com> writes:

>> Though I've since been thinking (and a quick codesearch[*] confirms
>> it) that we have a bunch of hostname==service/vhost assumptions
>> baked into our configuration management right now. While I'm not
>> opposed to the plan in principle, executing it implies a pretty
>> significant cleanup effort to stop using things like $::fqdn all
>> over the place. That isn't work that I _personally_ have bandwidth
>> to tackle right now, and it's a potentially disruptive effort for
>> each of the services whose modules/classes need this sort of
>> refactoring.
>> 
>
> At the summit Clark pointed out that many of our vhost templates could
> be simplified to listen on *. I think this works for http, not sure
> about https. I agree with the idea of getting a new host naming scheme,
> but I also agree that there will need to be some minor refactoring done.
> The problem, as you point out, is that we have $::fqdn sprinkled all
> over, so it would be a lot of churn to fix things everywhere. I would
> not support some kind of hack that overwrote fqdn.  I also don't think I
> have the bandwidth to refactor all of puppet right now.

Listening on * has the disadvantage of making colocating services more
difficult.  Of course, we have a lot of other things making that
difficult on accident since we generally don't do it.

-Jim



More information about the OpenStack-Infra mailing list