I'm having trouble with a Rails application on Heroku. The number of requests per minute it can handle has suddenly dropped SHARPLY.

I'm looking for a list of things to investigate for how to fix this, as we're releasing today and everything I've checked can't seem to increase this performance cap.

Specifically, my setup is: Heroku, Unicorn and Rails with a MongoHQ database.

Results 2 days ago: 40,000 requests per minute Results now: 2,000 requests per minute

I've tried the following steps to bring performance back up to previous levels, but none of them have made a difference and I'm stumped: - increase the number of dynos - swap to 2x dynos - roll back to the previous codebase and config files - disable Paper Trail logging - look for Heroku errors in the error logs - look for swapping or a large index in mongoHQ - check memory usage -- seems fine - check/adjust Unicorn settings -- more workers, shorter backlogs

The only change in requests per minute I've had success with is, lowering the dynos below 4 decreases the requests per minute.

What else haven't I checked?


You must log in to answer this question.

Browse other questions tagged .