After thinking about it a bit more, it would probably be much easier to
implement a module that takes as input a list of servers and outputs a
variable with the best available, which would then be used in the copy
statements.
This is the way I would look at doing it:
Lets say, for example, we have a bunch of buildings linked via a low
bandwidth link, so each building is on a different subnet and traffic
over the links is at a premium.
I would place a master server in each building, and each building would
have it's own dns server. The main building would have a primary
cfengine server, as well as a secondary (if it needed it on that site).
The DNS server on each site has its own local cfengine server as a
static (CNAME rather than A perhaps) entry, but it is the same name on
every site (but depending on which site you are on, you will get a
different IP address when you resolve the name).
Then in order to get the secondary servers to update from the primary,
you just set the secondaries to resolve to a master DNS server in the
main building, which resolves the server name to the primary cfengine
server.
Does this make sense?
Regards,
Marco.
_______________________________________________
Help-cfengine mailing list
[email protected]
http://lists.gnu.org/mailman/listinfo/help-cfengine
