[Varnish] #1086: VGZ_WrwGunzip loops forever if receiving junk data after end of gzip data
Varnish
varnish-bugs at varnish-cache.org
Tue Aug 27 22:50:24 CEST 2013
#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
More information about the varnish-bugs
mailing list