To accomplish this the second host in the chain needs to be a full version of splunk, not a Universal Forwarder.  This is because the intermediary will be acting as an indexer to collect any data forwarded to it. 
 So on your application server, in it's Universal forwarder instance, you will want an outputs.conf with something like: 
  [tcpout]
server=logforwarder
  
 Then, on the logforwarder machine, you will have a full splunk install, but it will also have a outputs.conf indicating where it should send it's data to: 
  [tcpout]
server=webpanel
  
 logforwarder does not need an inputs.conf, as it's not monitoring any logs directly.  It's simply accepting incoming data much like an indexer would.  You would also want to have any other props.conf stanzas present that are relevant at index-time.  (line breaking, timestamps, etc) 
						
					
					... View more