How can I be sure each process in my mongrel_cluster is handling requests?

Question

I have a rails application that takes a while to process certain requests (as it processes image uploads). I currently have three mongrel processes in a cluster, and I expect one of the other two to handle a second request if the first one is busy.

However, this doesn't seem to be happening. If I watch the output from top on the server, I can tell when it is resizing an image, so if I try to hit the application with an easy request (which would normally finish quickly), it seems to wait until the long-running request in finished.

When I look in the log/mongrel.xxxx.log files, all I see is the output from the application starting up.

How can I verify which process is actually handling each request?

Can I get them mongrels to log each request (even if it is just a timestamp, and something like 'GET /path/to/resource')?

Any ideas why the cluster might not be sharing the load in the first place?

The front-end web server is nginx, so maybe that is where I should be looking? The nginx access_log doesn't seem to have anything in it about where the request was proxied to.

Answer 1

We encountered similar issues on reevoo.com some time ago. Have a look at the resultant blog post.

Of specific interest is the use of mongrel_proctitle to diagnose the issue. It will allow a ps aux to see what a specific mongrel is handling.