Varnish behind load balancer not caching content

Question

I am using this line to remove cookies in default.vcl to allow for Varnish caching

set req.http.Cookie = regsuball(req.http.Cookie, "(^|;\s*)(__[a-z]+|__utma_a2a|has_js|Drupal.toolbar.collapsed|MNO_abc_qrst_\d+)=[^;]*", "");

but it does not appear to work.

Varnish is behind a load balancer, the load balancer sets a persistence cookie that appears in this format in the header:

Set-Cookie:MNO_abc_qrst_80=aaaaaaaaa2a5882045525d5a4a58455e445a4a423660;expires=Fri, 12-Aug-2011 17:23:23 GMT;path=/;httponly

I ran this regex in an emulator and it catches the above line (until the first ";"). So I'd think this regex should catch it, but it doesn't appear to? When I bypass this load balancer, content gets cached.

Any ideas? What am I missing?

Answer 1

There are 2 ways to go, either validate the entire line needs to be removed,
or cut out the offending parts of the line. I don't know what kind of regex engine you are using. At least negative look behinds and back references are necessary for a better result.

Compressed validation: /^(?=(Set-Cookie:\s*))(?:\1|.*?;\s*)(?:__[a-z]+|__utma_a2a|has_js|Drupal\.toolbar\.collapsed|MNO_abc_qrst_\d+)=[^;]*(?:;|$).*$/s

Compressed global part substitution: s/(?:(?<=^Set-Cookie:)|(?<=;))\s*(?:__[a-z]+|__utma_a2a|has_js|Drupal\.toolbar\.collapsed|MNO_abc_qrst_\d+)=[^;]*(?:;|$)//g

An expanded look at the regex's (in Perl):

my $str = 'Set-Cookie: MNO_abc_qrst_80=aaaaaaaaa2a5882045525d5a4a58455e445a4a423660;expires=Fri, 12-Aug-2011 17:23:23 GMT;path=/;httponly';

## Validate part of a cookie, remove line

if ( $str =~ /
       ^ (?= (Set-Cookie: \s*))
       (?:  \1
          | .*? ; \s*
       )
       (?:   __[a-z]+
          |  __utma_a2a
          |  has_js
          |  Drupal\.toolbar\.collapsed
          |  MNO_abc_qrst_\d+
       )
       =
       [^;]*
       (?: ; | $)
       .* $
 /sx )
{
     print "Valid, remove line '$&'\n=============\n\n";
}

## Globally, replace many parts of the cookie

if ( $str =~ s/
       (?:   (?<= ^ Set-Cookie:)
          |  (?<= ;)
       )
       \s* 
       (?:    __[a-z]+
           |  __utma_a2a
           |  has_js
           |  Drupal\.toolbar\.collapsed
           |  MNO_abc_qrst_\d+
       )
       =
       [^;]*
       (?: ; | $)
 //xg )
{
     print "removed parts of cookie\n";
     print "new string = '$str'\n";
}

Output:

Valid, remove line 'Set-Cookie: MNO_abc_qrst_80=aaaaaaaaa2a5882045525d5a4a58455e
445a4a423660;expires=Fri, 12-Aug-2011 17:23:23 GMT;path=/;httponly'
=============

removed parts of cookie
new string = 'Set-Cookie:expires=Fri, 12-Aug-2011 17:23:23 GMT;path=/;httponly'