Server-side web applications suffer from race conditions which cause data loss and incorrectness or even server crash. We studied bugs from popular server-side applicaitons written in PHP. We found that most web applications are performed in an event-driven way and client requests are handled by processes on the server-side. Most server-side races are on external resources such as files and database so existing client-side race detection tools cannot be applied. In this talk, I will present two previous works, Racepro and Node.fz. Racepro is a generic process race detection framework. It records live execution of the deployed system, detects races from analyzing the logs generated and validate if the detected races are harmful. Node.fz is a schedule fuzzing test aid tool that increases bug exposure rate. It modifies Node.js architecture to find schedule fuzzing opportunities. I will finish the talk by showing limitations of these two works on solving server-side race detection and our future plan.