Earlier this week I finally made the plunge to upgrade my VPS to Ubuntu 16.04. With a minor hiccup surrounding supervisord (which I probably can avoid if I go the systemd route) not being enabled at boot, the upgrade was simple for both my WSGI and Node webapps.

I can not say the same thing about my WordPress/PHP installations. (Installations that I hope to transition off to Rookeries once that software becomes more stable.) It took me a few hours to track down and resolve the problems. Hence I am posting this article, to hopefully save someone else’s time when they do the same upgrade.

Upgrading to PHP 7.0

Ubuntu 16.04 makes the switch away from PHP 5 to PHP 7. So I had to switch to php7.0, php7.0-fpm, and php7.0-mysql from their PHP 5 equivalents. The location of the running UNIX socket has changed from /var/run/php5-fpm.sock to /var/run/php/php7.0-fpm.sock, as did the PID files.

Updating the PHP-FPM configuration

Running WordPress using FPM (Fast Process Manager) and NGINX, requires turning off the path translation in php.ini file. This can be done by uncommenting the line cgi.fix_pathinfo=0 found in the configuration file /etc/php/7.0/fpm/php.ini. Again these files have moved from the old location. After you’ve done this remember to restart the FPM service using the new systemd utilities: sudo systemctl restart php7.0-fpm.

Updating the NGINX configuration and Solving the Blank Response

This is the tricky part. After updating my NGINX configurations to the new UNIX socket path, and restarting NGINX, I found that I got blank PHP responses. Everything else worked, expect that any PHP page would not render. And not render by not rendering any content in the body of the responses. That led me down a few rabbit holes, and researching how to re-architecture my setup using Docker. Eventually I stumbled across a blog entry with the solution to the blank PHP response issue.

In a nutshell, with the NGINX upgrade one of the parameters needed for FastCGI went missing namely the fastcgi_param SCRIPT_FILENAME $document_root$fastcgi_script_name; bit. Oddly this parameter appears in /etc/nginx/fastcgi.conf and not the /etc/nginx/fastcgi_params file that I normally include in my NGINX configs. Anyways after adding this line and restarting NGINX using sudo systemctl restart nginx everything worked correctly. Below I’ve included a sample NGINX configuration that should work.

Sample NGINX Configuration /etc/nginx/site-enabled/example-site.conf

“`
server {
server_name .example-site.com;
index index.php;
root /srv/www/location_of_wordpress_install;
listen 80;

access_log      /var/logs/nginx/site-access.log;
error_log       /var/logs/nginx/site-error.log info;

location / {
    try_files       $uri $uri/ /index.php$args;
}

location ~ \.php$ {                        
    fastcgi_split_path_info     ^(.+\.php)(/.+)$;
    fastcgi_pass    unix:/var/run/php/php7.0-fpm.sock;
    fastcgi_index   index.php;
    fastcgi_param   SCRIPT_FILENAME    $document_root$fastcgi_script_name;
    include         fastcgi_params;
}

location ~* \.(js|css|png|jpg|jpeg|gof|ico)$ {
    expires         max;
    log_not_found   off;
}

}
“`