A few days ago we received a notice from our hosting company, Dreamhost, that they'd be moving us to a new server.

That notice included a few hints & tips - basically a short list of things we might be doing wrong that would cause breakages after the move (e.g. wrong paths, hostnames fixed to machine names, etc.) There was also a profoundly rude message to PHP users, explaining that their apps might break (my emphasis):

Custom PHP
If you built your own custom PHP (or php.ini), it may be broken now. This is because
the old servers used a 32bit architecture and the new servers are 64bit. Unfortunately,
this isn't something we can help you troubleshoot, but you'll just need to rebuild your
custom php (or php.ini) again using the new servers libraries.

Not using custom PHP builds ourselves, we went through the rest of the list and checked we weren't doing anything wrong. Everything came up clean, so we waited patiently for the move.

This morning, two of our Rails apps (including the CRM we use for tracking sales data) failed to start. Closer examination revealed that the move had taken place overnight (AEST) and had indeed killed those apps stone dead. Both are straightforward Rails 2.x apps, with no special requirements.

I've logged a top-priority support ticket with DreamHost & await their reply. It'd better be damn helpful, or I'm going to start looking for another hosting company.

Update: Dreamhost got back to us in a few hours, during which time my colleague had realised that we were running some custom Gems that needed rebuilding to target the 64-bit architecture. This fixed the problem - and was exactly the solution suggested by the DH support guys.

I'm still not impressed by the downtime, but I am impressed by the support response time & usefulness.