Puppet » Facter

This is a really sweeping change. I am not even sure I concur with it. Thoughts Ken?

Doug Balmer wrote:

Some of us for various reasons like to have “hostname"s that are "relative names” that could contain a period. “example.com” below is the domain name.

So the premise here is that ‘hostname’ is really the relative name once the domainname is removed from fqdn. In some ways I concur as I have used multi-part ‘hostnames’ before like this but I think its very subjective and needs debate.

On the negative another way to look at this is that really a hostname could be relative to varying parts of a domain. So lets say you have:

foo1.production.city.country.mydomain.com

Out of the following what would the hostname be?

foo1
foo1.production
foo1.production.city
foo1.production.city.country

(and in-turn what is the domainname?)

I think the important question Doug is really what you want to do with this information that you can’t do without breaking down the $fqdn using functions like split etc. Or with say regexp matching for node definitions or using the $fqdn fact. What is the real world use case that makes this important to you?

After chatting to James I think I’m not wanting to make a decision on this one by myself. I would want a greater debate from the community I think as changing this known behaviour is dangerous for existing installations, and is to some extent the proposal has some subjective aspects.

I’m going to continue to watch this but remove myself as an assignee until something is agreed – but if you can state your case on the list Doug and reference this ticket that would be the best I think.

The user list is here:

http://groups.google.com/group/puppet-users

Thanks.

Current puppet-users discussion is here

So Nigel,

Right now the discussion has continued in the thread inidicated above and I think some good points have been made. I can’t get traction on priority as it seems only a couple of people are wanting this now but I’m of the believe that the solution proposed is not a terrible one to work around this issue. In that we create a set of new facts:

uname_nodename: is uname -n only and isn't contrived
uname_release: is uname -r
uname_version: is uname -v
...etc...

This duplicates a lot of facts in behaviour - but sticks to the posix
compliance interpretation only.

Dexter Plameras (easybeats – an old colleague of mine) has volunteered to help write a prototype. I’ll get him to join redmine and we can probably assign it to him if he still wants to help.

What do you think? I’m going to mark this as ‘low’ because I didn’t get a lot of people screaming for it – but its still important to some people – while it does have a purism feel about it – the solution does align more with the posix definitions.

+1 on all counts, apart from the actual naming….

does “uname_*” make sense for platforms like Windows?

woops. accidental assign to Thebo.

This was not an RC introduced bug. I’d classify as a feature request, asking for a new set of facts.

Assigning to Dexter Plameras as a volunteer contributor.

See CONTRIBUTING.md in the github repo for more information but if you need help Dexter, just ask the puppet-dev mailing list or myself.