Notes on beating the OSU configuration files into submission

Now that I have my own domain name (wyblog.us) naturally I want the web server on Ceres to serve the blog pages from that address, while still serving all my old pages (pictures, static pages, various pdf docs, etc) from their old URLs. It turns out that doing this wasn't intuitively obvious, and the online docs for the OSU web server didn't shed much light.

So, I embarked on some experimentation. The first config change was easy, adding a 2^nd ip address to Ceres. Tcpip$config.com makes this trivial, and the servermaint script allowed me to assign ceres.datalife.com to the old ip address and wyblog.us to the new one.

Then I encountered my first "feature". Any URL that worked on ceres.datalife.com also worked on wyblog.us. It took Googlebot about 5 minutes to start indexing all of my static pages using the new URL. My first reaction was, "boy those guys are fast", quickly followed by "how do I make it stop?".

The magic configuration commands for ip-based multi-homing are in www_root:[system]http_paths.conf. The servermaint script had put in the 2 ip addresses this way:

# Configure multi-homing (ip based and host/cname-based) root pages and
# log files.
#
.ITERATE localaddress cname $hname ;\
         AccessLog $cn_logfile $cn_extflags ;\
         map / $cn_root
.ITERATE localaddress $addr $name ;\
        AccessLog $mh_logfile $mh_extflags ;\
        map / $mh_root
.next 192.246.103.15 ceres.datalife.com access-ceres.log /
.next 192.246.103.16 wyblog.us access-wyblog.log 1 /blog
localaddress            # terminate localaddress blocks.

map /blog* /htbin/blosxom.pl*
map /wyblog* /htbin/blosxom.pl*

I thought this would map anything for wyblog.us into the Blosxom.pl script. Hah! It mapped the root page, but anything else fell through into the rest of my (gargantuation) path remapping maze. This is how Googlebot was able to start grabbing my static pages via the new URL.

I tried a whole bunch of different combinations of "map" trying to get things to go to the right place when I finally had an "aha!" moment. That "map /" in the localaddress directive, what if it was "map /*"? I was on to something for sure! A few tweaks later here's how I'm configured now:

.ITERATE localaddress $addr $name ;\
        AccessLog $mh_logfile $mh_extflags ;\
        map /* $mh_root
.next 192.246.103.15 ceres.datalife.com access-ceres.log /ceres/*
.next 192.246.103.16 wyblog.us access-wyblog.log 1 /blog/*
#
# now map the ceres and blog paths back
# this will prevent wyblog.us from serving anything but the blog and its images
map /blog/images/* /images/*
map /blog/blog* /blog*
map /ceres/* /*

map /blog* /htbin/blosxom.pl*
map /wyblog* /htbin/blosxom.pl*

This works like a charm. The trick is to take anything coming in for ceres.datalife.com and prepend "/ceres/" on to it. Then take anything coming in for wyblog.us and prepend "/blog/" on that. The only things other than blog posts that should be served by wyblog.us are the image files used by the blog templates. So I map all the blog stuff into the script rule for /blog and then map all the ceres stuff back to the root.

Now when wyblog.us gets a request for a page that ought to be served from Ceres it feeds the request into the Blosxom.pl script which doesn't know how to serve those requests so the requestor gets a blank page.

Posted at 13:23 by Chris Wysocki   [/vms]   Comments | Perm Link |   |   Tweet