[Varnish] #934: Memory leak in varnish_32e40a6ececf4a2ea65830e723c770d1ce261898

Varnish varnish-bugs at varnish-cache.org
Sat Jun 11 00:36:03 CEST 2011


#934: Memory leak in varnish_32e40a6ececf4a2ea65830e723c770d1ce261898
--------------------+-------------------------------------------------------
 Reporter:  kdajka  |        Type:  defect  
   Status:  new     |    Priority:  normal  
Milestone:          |   Component:  varnishd
  Version:  trunk   |    Severity:  normal  
 Keywords:          |  
--------------------+-------------------------------------------------------

Comment(by kb):

 varnishstat is confirming that Varnish is using exactly the 9GB you
 specified.  By "without any success" do you mean the memory usage remained
 unchanged?

 Run pmap on your child process.  256K allocations will be thread stacks (
 you can verify the option you're using is working ) and the 1024K
 allocations should be jemalloc blocks, and both will have matching single-
 page allocations (4K).  Then look for anything anomalous (and/or post it).

 Assuming invariant object size, causes of object bloat could be that you
 support Accept-Encoding:/Vary: which will create multiple version of
 objects.

 Do you do lots and lots of bans?  That's the other common memory user.
 Any inline C in your VCL?

 2.1.5 has been in production here since April 12 (example daemon still
 running with ~400 hits/s).  I haven't seen any leakage or bloat with
 2.1.5.  I expect that if this were endemic to 2.1.5, that a lot more
 people would be seeing this issue.

 Do you have any inline C in your VCL?

-- 
Ticket URL: <http://www.varnish-cache.org/trac/ticket/934#comment:3>
Varnish <http://varnish-cache.org/>
The Varnish HTTP Accelerator




More information about the varnish-bugs mailing list