This Play! Framework 2 app shows how to limit the concurrently processed requests in an async/non-blocking application. Requests exceeding this limit are queued and processed as soon as other requests are finished.
This may be useful in such a situation:
- Your play app is working async/non-blocking (e.g. by only using the WS library and/or asynchronous drivers)
- Sometimes the app receives very many requests in a very short time
- Processing all requests concurrently would produce so much data that the jvm would die with an OutOfMemoryError (gc overhead limit exceeded)
- A successful (expected) response is preferred over a rejected request if there are too many concurrent requests, even if it takes some "more" time until the response is available.
The solution (io.ino.play.ConcurrentRequestsLimiter) works with a limit of concurrently processed requests (configurable via
maxProcessedRequests, default 100), so that incoming requests, that would exceed this limit, are queued until other requests are completed.
Upon the completion of a request, the oldest queued request will then be processed.
Additionally, it's possible to limit the number of queued requests (configurable via maxQueuedRequests, default 10000).
Incoming requests that would exceed this limit are then in fact rejected with HTTP status 429 ("Too many requests").
The license is Apache 2.0, see LICENSE.txt.