[Varnish] #1540: Return pipe from vcl_backend_response

[Varnish]

#1540: Return pipe from vcl_backend_response
--------------------+-------------------------
 Reporter:  joakim  |       Type:  enhancement
   Status:  new     |   Priority:  normal
Milestone:          |  Component:  varnishd
  Version:  trunk   |   Severity:  normal
 Keywords:          |
--------------------+-------------------------
 Would it be possible to enable return (pipe) from vcl_backend_response. At
 the moment large, uncacheable, files can eat up all the memory on a
 machine by using transient storage when delivering to a client.

 In Varnish 3.0 it was possible to check the Content-Length and set a req
 header in vcl_fetch, then return (restart) and in vcl_recv do a return
 (pipe) if the req header is set. In Varnish 4.0 this doesn't seem possible
 due to the split nature of a request. When restarting in vcl_deliver the
 object is still being fetched into transient storage by the backend worker
 while a new pipe session is started for the restarted request. It would be
 nice to be able to switch to pipe mode immediately in vcl_backend_response
 to avoid out of memory conditions.

-- 
Ticket URL: <https://www.varnish-cache.org/trac/ticket/1540>
Varnish <https://varnish-cache.org/>
The Varnish HTTP Accelerator