William Warnecke

@Derevka @Moshne The hilarious part was people uploading 60+ days of raiding in one report. I had to push a new client just to stop that.

@KihraOfTemerity What do your front ends do? Could you be more resistant to high incoming data by doing little work except queueing?

@KihraOfTemerity Maybe what you're already doing and what was behind were workers to process queued items or queue lock contention?

@ww I fixed the stalled threads problem. WIll see how things go tonight. Hard to get a good read once the site falls behind.

@KihraOfTemerity At EA we used an API Front End -> RabbitMQ -> n Workers -> Database -> Web Front End. Worked well, easy to isolate issues.

@KihraOfTemerity Used that model for one project in particular that is. I fell in love with RabbitMQ. I'm an Erlang fan already though.

@ww Yeah ultimately I may do something like RabbitMQ (would just use Amazon's new service that is similar, since I'm on their platform).

@KihraOfTemerity That's particularly powerful if you can then scale the amount of instances running workers as needed to handle demand.