[Varnish] #1086: VGZ_WrwGunzip loops forever if receiving junk data after end of gzip data

[Varnish]

#1086: VGZ_WrwGunzip loops forever if receiving junk data after end of gzip data
----------------------+---------------------
 Reporter:  martin    |       Owner:  martin
     Type:  defect    |      Status:  new
 Priority:  normal    |   Milestone:
Component:  varnishd  |     Version:  3.0.2
 Severity:  normal    |  Resolution:
 Keywords:            |
----------------------+---------------------

Comment (by nnewton):

 I don't have much to add here, but wanted to note that I've also recently
 hit this with a cluster of 3 varnish 3.0.3 servers. It was pretty much
 identical to the above situations:
  * Varnish would run for about an hour and then one thread would take 100%
 of a single core, if you left it eventually another thread would as well.
  * If you pull the stack of these threads they are clearly looping through
 the gunzip process.
  * These events happened around incomplete transfers from the backend it
 would seem.
  * They were all on ESI-ed paths, thus forcing varnish to attempt to
 gunzip the object from the backend.

 We resolved the issue be editing the VCL to request un-gzipped objects
 from the backend on ESI-ed paths. This has solved the issue for us.

-- 
Ticket URL: <https://www.varnish-cache.org/trac/ticket/1086#comment:11>
Varnish <https://varnish-cache.org/>
The Varnish HTTP Accelerator