Hey all, I've been following the developments around AI and like what  
I see so far. I was paging through version 1 of the AI design doc and  
noted this section:

4.9.7 Host Name
In a controlled environment, each IP address has an associated host  
name. In those situation, the clients will know their hostname when  
they get the IP address. Setting up the installation based on hostname  
is easier for the administrator. So it makes sense to use hostname as  
the criteria to get the manifests.

This sounds pretty familiar, just like the rules file in Jumpstart.  
One thing that I've always desired in Jumpstart, and might be good to  
introduce here in AI, is the ability to use regular expressions,  
particularly for host names.

Say I have a ton of hosts, groups of which do specific jobs that are  
reflected in their hostnames... say like a bunch of web servers that  
are named like web1.foo.com, web2.foo.com, and so on. I might also  
have pools of database servers such as db1.blah.com, db2.blah.com,  
db3.blah.com....

It is reasonable to expect that the web servers and the db servers  
might get a similar basic OS load on them, but then take divergent  
paths regarding "begin" and "finish" scripts, extra packages and  
patches, configuration file tweaks and such.

Rather than list each and every host by name, with all the web servers  
having the same manifest, and all the db servers having their same  
manifest, it would be useful just to stick a regex in there.

Any hostname matching web\d+.foo.com gets the "webserver" manifest,  
for example.

Also, regarding AI criteria based on IP address range, would one be  
able to specify this range in CIDR notation?

TIA,
/dale

Reply via email to