Scripts for a WordPress Weblog Farm [en]

A first step to WordPress-farming: a shell script and a PHP script which allow you to easily install a whole lot of WordPress weblogs in only a few minutes (I installed over 30 in less than 5 minutes). Scripts require adapting to your environment, of course.

Update 03.11.06: Batiste made me realise I should point the many people landing here in the search of multi-user WordPress to WordPress MU. All that I describe in this post is very pretty, but nowadays completely obsolete.

Here is the best solution I’ve managed to come up with in half a day to finally install over 30 WordPress weblogs in under 5 minutes (once the preparation work was done).

A shell script copies the image of a WordPress install to multiple directories and installs them. A PHP script then changes a certain number of options and settings in each weblog. It can be used later to run as a “patch” on all installed weblog if a setting needs modifying globally.

Here are the details of what I did.

I first downloaded and unzipped WordPress into a directory.

wget http://wordpress.org/latest.tar.gz
tar -xzvf latest.tar.gz
mv wordpress wp-farm-image

I cleaned up the install (removing wp-comments-popup.php and the import*.php files, for example), added a language directory (as I’m wp-farming in French) and modified index.php to my liking; in particular, I edited the import statement for the stylesheet so that it looked like this:

@import url( http://edublogs.net/styles/toni/style.css );

The styles directory is a directory in which I place a bunch of WordPress styles. I don’t need the style switcher capability, but I do need to styles. Later, users will be able to change styles simply by editing that line in their index.php (or I can do it for them).

Another very important thing I did was rename wp-config-sample.php to config-sample and fill in the database and language information. I replaced wp_ by xxx_ so that I had $table_prefix = 'xxx_';.

To make it easier to install plugins for everyone, correct the language files, and edit whatever may be in wp-images, I moved these three directories out of the image install and replaced them with symbolic links, taking inspiration from Shelley’s method for installing multiple WordPress weblogs.

mv image/wp-content common
mv image/wp-images common
mv image/wp-includes/languages common
ln -s common/wp-content image/wp-content
ln -s common/wp-images image/wp-images
ln -s common/languages image/wp-includes/languages

I also added an .htaccess file (after some painful tweaking on a test install).

Once my image was ready, I compiled a list of all the users I had to open weblogs for (one username per line) in a file named names.txt, which I placed in the root directory all the weblog subfolders were going to go in.

I then ran this shell script (many thanks to all those of you who helped me with it — you saved my life):

for x in `cat names.txt`
do
cp -rv /home/edublogs/wp-farm/image/ $x
cat $x/wp-config.php | sed "s/xxx/${x}/" > config.tmp
mv config.tmp $x/wp-config.php
wget http://st-prex.edublogs.net/$x/wp-admin/install.php?step=1
wget http://st-prex.edublogs.net/$x/wp-admin/install.php?step=2
wget http://st-prex.edublogs.net/$x/wp-admin/install.php?step=3
done

This assumes that my WordPress install image was located in /home/edublogs/wp-farm/image/ and that the weblog addresses were of the form http://st-prex.edublogs.net/username/.

This script copies the image to a directory named after the user, edits wp-config to set the table prefix to the username, and then successively wgets the install URLs to save me from loading them all in my browser.

After this step, I had a bunch of installed but badly configured weblogs (amongst other things, as I short-circuited the form before the third install step, they all think their siteurl is example.com).

Entered the PHP patch which tweaks settings directly in the database. I spent some time with a test install and PHPMyAdmin to figure out which fields I wanted to change and which values I wanted to give them, but overall it wasn’t too complicated to do. You’ll certainly need to heavily edit this file before using it if you try and duplicate what I did, but the basic structure and queries should remain the same.

I edited the user list at the top of the file, loaded it in my browser, and within less than a few seconds all my weblogs were correctly configured. I’ll use modified versions of this script later on when I need to change settings for all the weblogs in one go (for example, if I want to quickly install a plugin for everyone).

In summary:

  1. compile list of users
  2. prepare image install
  3. run shell script
  4. run PHP script

If you try to do this, I suggest you start by putting only two users in your user list, and checking thoroughly that everything installs and works correctly before doing it for 30 users. I had to tweak the PHP script quite a bit until I had all my settings correctly configured.

Hope this can be useful to some!

Update 29.09.2005: WARNING! Hacking WordPress installs to build a farm like this one is neat, but it gets much less neat when your weblog farm is spammed with animal porn comments. You then realise (oh, horror!) that none of the anti-spam plugins work on your beautiful construction, so you weed them out by hand as you can, armed with many a MySQL query. And then the journalist steps in — because, frankly, “sex with dogs” on a school website is just too good to be true. And then you can spend half a day writing an angry reaction to the shitty badly-researched article.

My apologies for the bad language. Think of how you’re going to deal with spam beforehand when you’re setting up a school blog project.

Similar Posts:

Musings on a Multiblog WordPress [en]

Thinking about a solution to make WordPress MultiBlog. Comments, criticism and other ideas welcome — please join the fun. In particular, I bump into a hairy PHP include problem.

[fr] Je réfléchis à comment on pourrait donner à WordPress la capacité de gérer plusieurs blogs avec une installation. Je me heurte à un problème concernant les includes PHP. Feedback et autres idées bienvenues!

**Update June 2007: Try [WordPress Multi-User](http://mu.wordpress.org/) now.**

I’ve used Shelley’s instructions using soft links. I tried Rubén’s proof-of-concept, but got stuck somewhere in the middle.

So I started thinking: how can we go about making WordPress MultiBlog-capable? Here is a rough transcript of my thoughts (I’ve removed some of the dead ends and hesitations) in the hope that they might contribute to the general resolution of the problem. I have to point out my position here: somebody with a dedicated server who’s thinking of setting up a “WordPress weblog-farm” (for my pupils, mainly). So I’m aware that I’m not the “standard user” and that my solution is going to be impractical to many. But hey, let’s see where it leads, all the same. Actually, I think I probably reconstructed most of Rubén’s strategy here — but I’m not sure to what extent what I suggest differs from what he has done.

From a system point of view, we want to have a unique installation of WordPress, and duplication of only the files which are different from one blog to another (index.php, wp-config.php, wp-comments.php, wp-layout.css, to name a few obvious ones). The whole point being that when the isntall needs to be upgraded, it only has to be upgraded in one place. When a plugin is downloaded and installed, it only has to be done once for all weblogs — though it can of course be activated individually for each weblog.

From the point of view of the weblogs themselves, they need to appear to be in different domains/subdomains/folders/whatever. What I’m most interested in is different subdomains, so I’ll stick to that in my thinking. (Then somebody can come and tell me that my “solution” doesn’t work for subfolders, and here’s one that works for subfolders and subdomains, and we’ll all be happy, thankyouverymuch.) So, when I’m working with blog1.example.com all the addresses need to refer to that subdomain (blog1.example.com/wp-admin/, etc); ditto for blog2.example.com, blog3.example.com, blogn.example.com (I used to like maths in High School a lot).

As Rubén puts it, the problem with symbolic links (“soft links”) is called “soft link hell”: think of a great number of rubber bands stretched all over your server. Ugh. So let’s try to go in his direction, for a while. First, map all the subdomains to the same folder on the server. Let’s say blog1.example.com, blog2.example.com (etc.) all point to /home/bunny/www/wordpress/. Neat, huh? Not so. They will all use the same wp-config.php file, and hence all be the same weblog.

This is where Rubén’s idea comes in: include a file at the top of wp-config.php which:

  1. identifies which blog we are working with (in my case, by parsing $HTTP_HOST, for example — there might be a more elegant solution)
  2. “replaces” the files in the master installation directory by the files in a special “blog” directory, if they exist

The second point is the tricky one, of course. We’d probably have a subfolder per blog in wordpress/blogs: wordpress/blogs/blog1, wordpress/blogs/blog2, etc. The included file would match the subdomain string with the equivalent folder, check if the page it’s trying to retrieve exists in the folder, and if it does, include that one and stop processing the initial script after that. Another (maybe more elegant) option would be to do some Apache magic (I’m dreaming, no idea if it’s possible) to systematically check if a file is available in the subdirectory matching the subdomain before using the one in the master directory. Anybody know if this is feasible?

The problem I see is with includes. We have (at least) three types of include calls:

  • include (ABSPATH . 'wp-comments.php');
  • require ('./wp-blog-header.php');
  • require_once(dirname(__FILE__).'/' . '/wp-config.php');

As far as I see it, they’ll all break if the calling include is in /home/bunny/www/wordpress/blogs/blog1 and the file to be called is in /home/bunny/www/wordpress. What is wrong with relative includes? Oh, they would break too. Dammit.

We would need some intelligence to determine if the file to be included or called exists in the subdirectory or not, and magically adapt the include call to point to the “right” file. I suspect this could be done, but would require modifying all (at least, a lot of) the include/requires in WordPress.

Maybe another path to explore would be to create a table in the database to keep track of existing blogs, and of the files that need to be “overridden” for each blog. But again, I suspect that would mean recoding all the includes in WordPress.

Another problem would be .htaccess. Apache would be retrieving the same .htaccess for all subdomains, and that happens before PHP comes into play, if I’m not mistaken.

Any bright ideas to get us out of this fix? Alternate solutions? Comments? Things I missed or got wrong? The comments and trackbacks are yours. Thanks for your attention.

Similar Posts: