I am running Rail 2.3 on bamboo-mri-1.9.2 stack on Heroku. However I am encountering some scaling issue. During peak hours of my traffic, I am seeing some Error H11 (Backlog too deep) and I can see that via New Relic monitoring that I am having quit a high back log. However, what I would like to know now, is not the length of the queue, but the number of incoming request that are rejected by Heroku (which would result in "Error H11 (Backlog too deep)" in the Heroku log).
In short, I would like to know the ratio of incoming requests being reject to the total number of incoming requests.
As Heroku said, the log file is mainly used for debugging purpose but not for analysing real live metric, is there easy/free way to do that?
Thanks!
P.S. Comment on this post How to keep log tail alive on Heroku using ssh? suggested using the papertrailapp.com website (not related to the papertrail gem apparently), but I couldn't determine whether it can give me indication of the ratio of H11 backlog too deep error.
There is no way to see the number of rejections as far as I am aware as they are turned away by the routing mesh - long before you see the request near your app.
More importantly, why are you not scaling the dynos to handle the traffic and not be turning away users with errors?