Staging magento sites: multiple sites running APC == broken

Well, it’s been a wild webhosting ride recently.  I’ve been automating dev/staging/live production procedures with my Drupal and Magento clients.  In a few days I’ll publish a comprehensive blogpost about this whole project, replete with nginx, php, php-fpm, etc configuration goodies.

For now, here’s one tidbit that I just can’t hold back:

I ran into an interesting problem with running multiple Magento sites on the same server: the new Magento devsite I set up would unfailingly read the old site’s db configuration information (from app/etc/local.xml) 9/10 times after restarting nginx and php5-fpm. And to make things even more confusing, on some restarts, it’d read the devsite db configuration and so would the live site!

After trying every single possible iteration of php5-fpm and nginx options that I thought might be relevant, I remembered that I had set up Magento to use APC for caching, and changed the prefix:

1
2
3
4
5
6
7
8
9
10

<config>
  <global> 
<!-- ... -->
        <cache>
            <backend>apc</backend>
            <prefix>alphanumeric (i.e. change this to something else)</prefix>
        </cache>
<!-- ... -->
     </global>
  </config>

It took a LONG time to track down this issue.  I had forgotten that I had apc cache running on this site, and I never would’ve guessed that it would mix up different site’s caches.  The search was even more complicated because I am not as confident with my nginx/php-fpm skills as i am with apache/mod_php. Serves me right for blindly implementing the magento apc cache without understanding what the various xml config options mean.

After finding the problem on my own, I searched to see if anyone else has experienced this and found a post on the magento board.  Wish I’d seen that 6 hours ago.

I’m going to add an APC-prefix-changing sed operation to my magento dev/staging/production scripts now.