Out-of-memory errors are a common JVM production problem. Finding the real cause means understanding what the JVM has retained in the heap, why it is still reachable, and which objects are keeping your memory alive.

In this talk, we’ll work through real OOM examples and learn how to turn a heap dump into a root cause. We’ll look at the HPROF format, analyze dumps with Eclipse MAT and hprof-analyzer, and explore where AI can help with heap analysis and where it cannot.

We’ll also look at practical ways to capture heap dumps from production, from `jcmd` to integrations with Spring Boot and Cloud Foundry.

But there’s a catch: a heap dump is effectively a snapshot of your application’s memory, and that memory may contain credentials, personal data, database contents, and other sensitive information. We’ll see how easily that data can end up in a dump, why sharing production dumps can be risky, and how `hprof-analyzer redact` can sanitize them before they leave the production environment.

By the end, you’ll know how to go from an `OutOfMemoryError` to the actual problem and a fix, without leaking your production data.