Rails 3.0, Windows, Apache, Thin > production stops working

Using the following guide I was able to get the newer version of rails
(3.0.x) running on my Windows production server:

Everything seems to be working and at first it seems faster than the old
Mongrel cluster.

I run 3 thin servers and I am using the ProxyPass directive to pipe
requests into a loadbalancer cluster.

It all seems to be running fine, except after a random time (usually
days) all 3 thin servers stop responding, causing the apache server to
throw a 503: service unavailable.

When I go into the server, I see the Ruby.exe processes still running,
but even navigating to http://localhost:4003 (and 4004 and 4005) does
nothing and gives a timeout.

Does anyone have an idea how to solve this issue?

I am using:
Ruby 1.8.7
Rails 3.0.10
Thin 1.3.1
Windows Server 2008 R2

What DB are you using? MySQL? Do you have keep alive connections or attempt to reconnect?

For long established connections I’ve encountered problems with MySQL and Ruby using mysql adapter. It required me set “reconnect: true” or specify some timeout mechanism to certain services (Redis for example).

Is there a way you can obtain a log from one of those thin installation behind Apache?