[Varnish] #1141: VCL's should stop polling backends on being discarded

[Varnish]

#1141: VCL's should stop polling backends on being discarded
----------------------+-----------------------------------------------------
 Reporter:  nicholas  |        Type:  defect
   Status:  new       |    Priority:  normal
Milestone:            |   Component:  build 
  Version:  3.0.2     |    Severity:  normal
 Keywords:            |  
----------------------+-----------------------------------------------------
 After removing a backend by vcl reloadand discard, we still see health
 checks of backend.

 A typical vcl.list looks like this 30 minutes after discard:
 discarded     203 boot
 active         54 1337845014

 The backend doesn't get used after discard, so this only affects backend
 health checks. Our monitoring of changes and application health is based
 around "varnishlog -u -i Backend_health". This hampers our debugging so
 much that we restart varnish on deploying new servers.

 Steps to reproduce: drop backend in vcl, reload vcl, discard vcl. Migth
 depend on varnish server being used actively, so there are threads which
 have been used, but are not currently in use.

 Greetings
 Nicholas

-- 
Ticket URL: <https://www.varnish-cache.org/trac/ticket/1141>
Varnish <https://varnish-cache.org/>
The Varnish HTTP Accelerator