More Detailed Logging (varnishncsa) and Clustering/Cache sharing questions

Javier Frias jfrias at
Mon Mar 9 22:17:46 CET 2009


congrats on a great product. So far in my testing, it has met and exceed my
expectations. A fresh change from the squid days :)

I'm still on the testing phase of my first varnish deployment, and have come
up with a few questions.

varnishncsa currently only seems to log on a fix format,  %h %l %u %t "%r"
%>s %b "%{Referer}i" "%{User-agent}i", while the code and man page mention
that the availability of a custom format will be forthcoming, has anyone
patched it to add things like bytes served and request time or a custom
header? Or can anyone answer if the varnish in memory log has these

Also, there are hints of clustering support in the wiki. I'm specifically
interested in cache sharing between varnish installs in a cluster. ( akin to
the Digest feature of Squid, where squid servers constantly share what they
are caching with each other ). Is there any initial support for this, or is
this planned in the near future?

And lastly, how stable is the backend load balancing? I have dedicated
hardware load balancers for my application, and have been debating whether
to point the varnish caches to it, or list all backends on varnish. The
plusses for listing them in varnish, are of course, having more detailed
tests, and retry logic, but just curious what other high traffic sites have

Again, thanks for a great project.

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <>

More information about the varnish-dev mailing list