How to use properly exported resources with Puppet - puppet

Question

I've been thinking for hours on how to solve a problem with Puppet 4.
Here is my case :
I have a module "Cassandra", and I have 3 machines
Cluster1 (Hostname : CassandraCluster1)
Cluster2 (Hostname : CassandraCluster1)
Cluster3 (Hostname : CassandraCluster1)
I want to collect the hostnames of three clusters in an array, so I can pass them to the cassandra configuration file (which I'm using as a template epp) :
cassandra.yaml.epp
- seeds: "<%= $cluster::hostnames %>"
So the solution is Exported Resources I came up with :)
I've been playing with all day, but no idea how to make up this work. here is what I've tried :
on each cluster I add this code to collect the hostnames :
##file {"${hostname}":
content => 'epp(puppet://modules/cassandra/cassandra.yaml.epp)',
}
# Collect:
File <<| |>>
But I'm not sure if this is a good idea ?!

Answer 1

I've found a temporary solution that I'm using now :
I have a role fact which I'm adding to each machine. and I'm retrieving the hosts from puppetdb using puppetdbquery module.
and then on puppet code I use this query to find the nodes with the role attached
$hosts = query_nodes("role=apache")

Answer 2

To the extent that the question is how to create the wanted file with use of exported resources, it's important to understand that collecting exported resources causes them to be added to the target node's catalog. Exporting and collecting resources is not merely a data-transfer excercise. Your example exports several File resources; collecting these will result in the same number of separate files being managed on the target (or else a duplicate resource error if you happen to have a resource title collision).
For a long time, the usual way to build a file based on pieces contributed by multiple nodes was to use the puppetlabs/concat module, with each contributing node exporting a concat::fragment resource. For example:
##concat::fragment { "${hostname} Cassandra seed":
target => '/path/to/file',
content => " ${hostname}",
order => '10',
tag => 'cassandra'
}
Those would be collected into the catalog for the target node, to be used in conjunction with a corresponding concat resource declared there:
concat { '/path/to/file':
ensure => 'present',
# ...
}
concat::fragment { "Cassandra seed start":
target => '/path/to/file',
content => ' - seeds: "'
order => '05',
}
Concat::Fragment <<| tag == 'cassandra' |>>
concat::fragment { "Cassandra seed tail":
target => '/path/to/file',
content => '"'
order => '15',
}
That still works just fine, but it is becoming more common to rely instead on querying puppetdb, as you demonstrate in your own answer.