Memory leaks are difficult to trace down. A nice description of finding one is "A server memory leak", a story about tracing a memory leak in a Django application.

I'm trying to work out why a server process is slowly using more and more memory. The server parses a stream of text, and shoves stuff into a database. It's written in Python, and about the only external code it uses is SQLAlchemy

I found a combination of Meliae and objgraph useful. Pympler also seems useful (the tracker class particularly). Dowser also looks useful, and might have made things a bit easier..

A common problem we all have to deal with is fixing memory leaks. In garbage collecting languages like Java we typically expect not to ever need to worry about memory management. However the limitations of how a garbage collector works means that we can still create leaks. Which defeats the advantage of a garbage collector.

How do leaks get formed?

The Java GC determines which objects to collect by looking for references to the object. If there is no references, then the object is no longer in use and can be safely destroyed without making the application unstable. If an object is now unused but still referenced, a leak occurs. Let's create a leak: