hanging connections

[Kristian Lyngstol]

On Tue, Dec 21, 2010 at 09:36:14AM +0100, Stig Bakken wrote:
> Kristian, please let me know whether I should file a ticket for this.

So far I haven't seen anything that suggests a bug.

A single slow response is all the logs show - for better or worse. That
could just as well just be a slow client or packet loss.

What I need to be able to determine if this is a bug in Varnish,
network-issue or phase-of-moon glitch is:

- Several occurrences of slow responses and the requests around it in
  varnishlog (you can obfuscate the IP - but I need to be able to tell
  which requests come from the same IP: ie: if there are two IP's, A.A.A.A
  and B.B.B.B are ok, but X.X.X.X for both are not).
- 'varnishstat -1' output
- 'sysctl -a' output
- Network configuration on the machine(s), specially with regards to any
  load balancers
- 'dmesg'
- Complete VCL (I do believe I already have that)
- All start-up parameters. Any other running services on the machines.

If that doesn't reveal the nature issue, I'll need to see tcpdump-output of
the incident to track it further - at that point, obfuscation is not
practically possible without risking the integrity of the data.

In short: all you can give me.

The issue doesn't have an obvious explanation and there isn't really any
reason why a cache hit would take any real amount of time when we rule out
busy objects. I should also point out that I can only really spend time on
this to rule out or confirm if it is a varnish bug or not.

-- 
Kristian Lyngstøl
Product Specialist
Varnish Software
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 490 bytes
Desc: Digital signature
URL: <http://www.varnish-cache.org/lists/pipermail/varnish-misc/attachments/20101221/1c30bc5f/attachment.pgp>