We have a 5 node cluster and ganglia server will not start. It seems to be that the startup script does not like that we changed the location of ganglia during install. Instead of /var/lib/ganglia we set it as /var/hadoop/lib/ganglia. It seems Ambari created a sym link to the /var/lib/ganglia directory and this is causing an error with starting rrdcached. When we updated the path in the start script ganglia server started back up but it seems puppet reverted the change back and now we cannot start ganglia server again. Where would we change the puppet config to update the rrdcached start script so that it will reflect the /var/hadoop/lib/ganglia directory directly?