What is in a girl’s head?

In our bathroom we have a framed poem on the shelf by the toothbrushes.

The poem is ‘A boy’s head’ by the late Czech poet and scientist Miroslav Holub

In it there is a spaceship
and a project
for doing away with piano lessons.

And there is
Noah’s ark,
which shall be first.

And there is
an entirely new bird,
and entirely new hare,
an entirely new bumble-bee.

There is a river
that flows upwards.

There is a multiplication table.

There is anti-matter.

And it just cannot be trimmed.
I believe that only what cannot be trimmed
is a head.

There is much promise in the circumstance that so many people have heads.

Miroslav Holub

I often read this poem while I brush my teeth.

It always make me wonder what might be in a girl’s head.

I have some theories, but am still working on a definitive list…


Keep an eye on your MongoDB

I don’t use MongoDB that often, but sometimes it is the perfect tool for a job.

I have one database with about 2 million documents in the main collection. We needed to rebuild and reprocess the documents in the collection recently – essentially emptying the collection and letting it rebuild naturally as new data came in.

After a while I started to notice some severe performance issues on the server as the collection grew. Mongo was consuming vast amounts of CPU. Eventually the penny dropped. After the update, the new indexes had not been created.

As soon as the indexes were added, the amount of resource used by Mongo dropped back to an almost irrelevant level. I am astonished as the performance difference they made.

Not a mistake to make again!

phpize error on RHEL server

Trying to install the MongoDB PHP driver using Pecl (as per the documentation) on a Rackspace managed server running RHEL I ran into the following problem:

pecl install mongo
/usr/bin/phpize: /tmp/tmpNW0rIa/mongo-1.2.12/build/shtool: /bin/sh: bad interpreter: Permission denied
Cannot find autoconf. Please check your autoconf installation and the
$PHP_AUTOCONF environment variable. Then, rerun this script.
ERROR: `phpize' failed

It took me a while to remember and then search through old tickets, but the problem is that on Rackspace’s RHEL servers the /tmp partition is mounted without script execute permissions. (Is this standard for RHEL in general?)

Pecl uses the /tmp – and so therefore, the install fails.

The solution is to temporarily allow script execution in /tmp

[root@server ~]# mount | grep /tmp
/dev/sda2 on /tmp type ext3 (rw,noexec,nosuid,nodev)
[root@server ~]# mount -o remount,exec /tmp
[root@server ~]# mount | grep /tmp
/dev/sda2 on /tmp type ext3 (rw,nosuid,nodev)

Install the MongoDB driver with Pecl…

[root@server ~]# pecl install mongo

Revert the permissions on /tmp and then check the permissions are correct…

[root@server ~]# mount -o remount,noexec /tmp
[root@server ~]# mount | grep /tmp
/dev/sda2 on /tmp type ext3 (rw,noexec,nosuid,nodev)

Caching Minified files with Nginx

A quick introduction to Minify

Minify is a popular PHP5 library best described by the Minify site itself:

Minify is a PHP5 app that helps you follow several of Yahoo!’s Rules for High Performance Web Sites.
It combines multiple CSS or Javascript files, removes unnecessary whitespace and comments, and serves them with gzip encoding and optimal client-side cache headers.

The uses of Minify are twofold:

  1. Speed up websites by combining and minifying files
  2. Help eliminate the problems of cached css and javascript files in users browsers.

Minify is fairly easy to setup and can dramatically increase the performance of websites, but as mentioned in the FAQs because each request is served by PHP, it can actually slow your site down (for example if your site recieves a lot of traffic or you have are on a feeble shared server)

A simple solution to this problem – suggested in the Minify FAQs  is to serve your Minified files through a reverse proxy.

An even quicker introduction to Nginx

Nginx is popular high performance web server and reverse proxy server. I’m not going to try and summarise what it does here – if you are reading this, the chances are you already know.

We have used Nginx for a while in a standard way, using Nginx in front of Apache. This approach is easy to configure and is also to revert to a pure Apache setup. We served all static content (css, js, images etc.) directly from Nginx and passing requests for dynamic content to a backend (PHP on an Apache server) using the Nginx HttpProxyModule.

Despite the fact that our Minified css and Javascript rarely change with this setup a request for either is still a PHP request. We can avoid this request by using the Nginx HttpProxyModule  to cache responses from the backend making a faster response and reducing the load on the backend server.

Nginx configuration

This is a simple Nginx configuration taken from a development server running Ubuntu. For ease of maintenance, the configuration file is split up into several files using sites-available and sites-enabled directories (Debian/Ubuntu style). In this case Nginx is listening on Port 80 and Apache is listening on Port 8080


The main file is nginx.conf which then includes all the live virtual hosts that are in the sites-enabled directory (often organised symbolic links to actual files in the sites-available directory)

The white-space in the config files doesn’t matter so it makes sense to organise your .conf files in the way you find most readable.

worker_processes  1;
events {
    worker_connections  1024;
http {
    include mime.types;
    default_type application/octet-stream;
    log_format  new_log
    '$remote_addr - $remote_user [$time_local] $request '
    '"$status" $body_bytes_sent "$http_referer" '
    '"$http_user_agent" "$http_x_forwarded_for"';
    # Proxy cache and temp configuration.
    proxy_cache_path 	/var/www/nginx_cache levels=1:2
			max_size=1g inactive=30m;
    proxy_temp_path 	/var/www/nginx_temp;
    sendfile on;
    include /etc/nginx/sites-enabled/*; 

The key directive here are the lines that setup the Proxy cache and which allow us to cache the results of scripts that are returned by the backend server. (see documentation)

    # Proxy cache and temp configuration.
    proxy_cache_path 	/var/www/nginx_cache levels=1:2
			max_size=1g inactive=30m;
    proxy_temp_path 	/var/www/nginx_temp;

proxy_cache_path – where on the filesystem the cached files will be stored
levels – helps define the structure of files stored in the cache directory
max_size – the maximum size of the cache
inactive – timeout period for requests to the cached files
proxy_temp_path – a buffer for requests from the file system


This is a simple example of one of the virtual host .conf files this is included using the include /etc/nginx/sites-enabled/*; directive in the nginx.conf file.

server {
    listen 80;
    access_log /var/log/nginx/www.example.com.access.log;
    error_log /var/log/nginx/www.example.com.error.log;
    root /var/www/www.example.com/;
    index index.php index.html;
    server_name www.example.com;
    # send appropriate headers to enable browser caching for static files
    # static files are identified by file extension
    location ~* ^.+.(jpg|jpeg|gif|png|ico|css|zip|tgz|gz|rar|bz2|doc|xls|exe|pdf|ppt|txt|tar|mid|midi|wav|bmp|rtf|js)${
    access_log off;
    expires 30d;
    # Set the proxy cache key
    set $cache_key $scheme$host$uri$is_args$args;
    location ~/min/ {
        # Set proxy headers for the passthrough
        proxy_set_header Host $host;
        proxy_set_header X-Real-IP $remote_addr;
        proxy_set_header X-Forwarded-For $proxy_add_x_forwarded_for;
        proxy_cache main;
        proxy_cache_key $cache_key;
        proxy_cache_valid 30m; # 200, 301 and 302 will be cached.
        # Fallback to stale cache on certain errors.
        # 503 is deliberately missing, if we're down for maintenance
        # we want the page to display.
        proxy_cache_use_stale 	error
    # proxy any other requests back to the Apache server listening on Port 8080
    location / {
        more_clear_headers 'Content-Length' 'Transfer-Encoding';
	proxy_cache_bypass 1;
	proxy_no_cache 1;
        proxy_set_header X-Real-IP $remote_addr;
        proxy_set_header X-Forwarded-For $proxy_add_x_forwarded_for;
        proxy_set_header Host $host;

The virtual host .conf file above contains 3 different location blocks each with a different role.

The first location block catches requests for static files and serves them directly from the file system – in this case the Apache server is not accessed at all. Unfortunately request to minify are all .php requests so cache headers are not sent an every time a minified file is requested it will be passed back to Apache.

The role of the third location block is to send any request that hasn’t already been dealt with already back to the Apache server.

The role of the second location block is deal with any requests to minify. We have set a cache timeout of 30 minutes and any requests within that time period will be served from the Nginx cache – if there is no match in the cache, the request will be passed back to Apache. If you are using minify_group, when the underlying Javascript and CSS files change, the timestamp on the minified URL will change and therefore no matching cache file will exist, so the Nginx cache will update.

How do I know if it’s working

The easiest way to see this is working is if you have Apache mod_status enabled. You will know it works because you will no longer see any requests to /min/?