Varnish VCL: Delete ALL cookies and other magic

This morning Javi Call√≥n gave me a great introduction in few minutes to the Varnish in steroids world, I really appreciate it.  I'd like to share this snippet which might be very interesting for you if you are new to the Varnish magic too.

This has been my first contact with Varnish ever, and I have to say I am quite amazed on how the application is responding now in terms of performance. Do not take this snippet as a definitive solution to your problems.

I wrote an article yesterday on how to install Varnish. Truth is that if you install Varnish and you do not tune the VCL file chances are that Varnish is not caching anything because of the cookies. In a dynamic application there are a lot of factors that have to be taken (headers, user-agents, variations...)

This sample VCL tries to address the following problems:

Open the file/etc/varnish/default.vcl and add the following. It is recommended to leave the rest of the comments as they are for future reference of what Varnish does by default.

The code is commented so you properly understand what it does and you can remove any pieces you don't need.

# You already have a block like this one when you installed Varnish, keep it safe:
backend default {
  .host = "127.0.0.1";
  .port = "8080";
}

# ADD THE FOLLOWING
# -----------------

# 2 things are done here:
# First, ignore any request to a specific host. For instance, you don't want Varnish on a specific host.
# Second, remove cookies, because my application does not rely on cookies at all.
sub vcl_recv {

     # Varnish will Ignore any request to this host  (e.g: xx.mydomain.com)
     if ( req.http.host ~ "([a-z0-9]{2}\.mydomain\.com)$" )
     {
        return(pipe);
     }

     #Goodbye incoming cookies:
     unset req.http.Cookie;

}


sub vcl_fetch {
    # Remove cookies that destroy cache:
     unset beresp.http.Set-Cookie;

     # 5 minutes (300s) cache for images
    if ( req.url ~ "\.(jpg|jpeg|png|gif)$" )
     {
        set beresp.ttl = 300 s;
     }

    # This is very specific of SIFO.me framework, but you can recycle it:
    # Any static URL containing ?rev= (this is JS and CSS) cache it almost forever.
    # The following regexp will find urls like http://.../file.js?rev=1747c3872495221156287e2000a0d110
    if ( req.url ~ "\?rev=[a-f0-9]{32}$" )
    {
       set beresp.ttl = 600000 s;
    }
}

# Add some debug info headers when delivering the content:
# X-Cache: if content was served from Varnish or not
# X-Cache-Hits: Number of times the cached page was served
sub vcl_deliver {

        # Was a HIT or a MISS?
        if ( obj.hits > 0 )
        {
                set resp.http.X-Cache = "HIT";
        }
        else
        {
                set resp.http.X-Cache = "MISS";
        }

        # And add the number of hits in the header:
        set resp.http.X-Cache-Hits = obj.hits;
}

Try if the configuration syntax is OK with:

/etc/init.d/varnish configtest

And then restart the service (also wipes the cache).

/etc/init.d/varnish restart

The new headers should appear and you will be able to see what Varnish is doing with a simple CURL or with any browser Inspector. Example:

[root@mnm1 mnm]# curl --head http://yourhost,com
HTTP/1.1 200 OK
Server: Apache/2.2.15 (CentOS)
X-Powered-By: PHP/5.3.3
Cache-Control: public, must-revalidate, max-age=30, s-maxage=43200
Vary: Accept-Encoding
Content-Type: text/html; charset=UTF-8
Date: Tue, 09 Oct 2012 13:38:41 GMT
X-Varnish: 1572155458 1572155457
Age: 12
Via: 1.1 varnish
Connection: keep-alive
X-Cache: HIT
X-Cache-Hits: 1

Have in mind that all these functions are appended to the default behaviour. So they are adding extra things, but not preventing the default Varnish workflow take action.

In the cases where the time to live of the cache (TTL) is not set ,Varnish will cache it for 2 minutes (look for "120 s" in the deafault.vcl code).

All in all, it seems to me that for the huge benefit that Varnish adds to a project, the investment of time and resources you have to put on are ridiculous (half morning if your dynamic app doesn't have excessive magic). Do not be scared and try to add Varnish at least in your static files, then move to the dynamics.