Ball Dawg!

When I started in the “Technology” department at my current employer, I found myself apart of a team that was tasked with taking care of hundreds of IBM blade servers, and tens of other IBM system x servers.  For the most part we could keep up with our servers by where they were in our monitoring software, but if we needed to know exactly where they were in either a blade center, by remote console name, or what Domain-0 they lived on for our Xen based virtual machines – we had to relate back to a usually out of date spreadsheet that showed where to go.

Read more…

I recently switched this server to Lighttpd (using PHP through FastCGI) from Apache. It was easy enough and I ended up with much faster serving websites, unfortunately I hit a snag on one of my sites that extensively uses .htaccess file for rewrites – which Lighttpd does not support. I didn’t want to bail on the whole switch because of a single site, so looked and came up with a simple solution – proxy to Apache through Lighttpd for items on that site (www.350z.ws). In lighttpds config, this was very easy to accomplish:

# Proxy 350z.ws back to apache
$HTTP["host"] =~ "www.3(5|7)0z.ws" {
        $HTTP["url"] !~ "(wp-content|wp-includes|css|js|php$|^/blog/$)" {
                proxy.server = ( "/" =>
                        ( "localhost" =>
                                ( "host" => "127.0.0.1", "port" => 81 )
                        )
                )
        }
}

Note one (obvious) draw back is that you have to run Apache as well, but since I’m stopping most hits at Lighttpd by serving everything in wp-content, wp-includes, anything with css, js or php in the name, I can greatly reduce the number of Apaches I start and maintain. In my case my Apache prefork config looks like this:

<IfModule mpm_prefork_module>
    StartServers          3
    MinSpareServers       2
    MaxSpareServers       5
    MaxClients          15
    MaxRequestsPerChild   10
</IfModule>

You’ll also notice I’m not proxying the folder where WordPress lives back either, this is because it contains no rewrites for it specifically and will get the majority of the hits.

My employer starting using lighttpd on one layer of our architecture about a year or so ago, until now that layer has kind of been a black box to the majority of the technical staff due to not having mod_status enabled.  In preparation for it being turned on (I requested it be so after using it on my own servers), I have created a Xymon Monitor (formally know as Hobbit) script which hits the /server-status page on the localhost and reports that data back to Xymon.  The data it reports includes requests per second and “amount increase since last script run” for the “Total KBytes” and “Total Accesses” numbers.  I also created a graph for the requests per seconds stat.

The Graph definition is as follows:

[lighttpd]
TITLE lighttpd Requests/Second
YAXIS # reqs/sec
DEF:RPS=lighttpd.rrd:reqpersec:AVERAGE
LINE2:RPS#0000CC:reqs/sec
COMMENT:
GPRINT:RPS:LAST:Requests per Second   : %5.1lf (cur)
GPRINT:RPS:MAX: : %5.1lf (max)
GPRINT:RPS:MIN: : %5.1lf (min)
GPRINT:RPS:AVERAGE: : %5.1lf (avg)

Read more…