Tune performance > Obtain advice from the advisors > Use the Performance and Diagnostic Advisor > Start the lightweight memory leak detectionLightweight memory leak detection
This topic describes memory leaks in Java applications and introduces lightweight memory leak detection.
Although a Java application has a built-in garbage page mechanism, which frees the programmer from any explicit object deallocation responsibilities, memory leaks are still common in Java applications. Memory leaks occur in Java applications when unintentional references are made to unused objects. This occurrence prevents garbage page from freeing memory.
The term memory leak is overused; a memory leak refers to a memory misuse or mismanagement. Old unused data structures might have outstanding references but are never garbage collected. A data structure might have unbounded growth or there might not be enough memory that is allocated to efficiently run a set of applications.
Most existing memory leak technologies are based upon the idea that we know that you have a memory leak and want to find it. Because of these analysis requirements, these technologies have significant performance burdens and are not designed for use as a detection mechanism in production. This limitation means that memory leaks are generally not detected until the problem is critical; the application passes all system tests and is put in production, but it crashes and nobody knows why.
WebSphere Application Server has implemented a lightweight memory leak detection mechanism that runs within the WebSphere Performance and Diagnostic Advisor framework. This mechanism is designed to provide early detection of memory problems in test and production environments. This framework is not designed to provide analysis of the source of the problem, but rather to provide notification and help generating the information required to use analysis tools. The mechanism only detects memory leaks in the Java heap and does not detect native leaks.
The lightweight memory leak detection in WAS does not require any additional agents. The detection relies on algorithms that are based on information that is available from the Performance Monitoring Infrastructure service and has minimal performance overhead.