How did you take heap dumps, analyze them, and automate the process?

Heap dump is a snapshot of the JVM heap at a specific moment, which helps analyze memory usage and detect leaks.

How to take a heap dump:

• Using built-in JVM tools, for example, jmap -dump:live,format=b,file=heapdump.hprof <pid>.
• In production, you can trigger a heap dump via JMX or using jcmd <pid> GC.heap_dump <file>.
• Sometimes, heap dumps are created during OutOfMemoryError by configuring the JVM with the -XX:+HeapDumpOnOutOfMemoryError parameter.

How to analyze:

• Opening the heap dump in tools like Eclipse MAT (Memory Analyzer Tool) or VisualVM.
• Looking for suspicious objects that occupy a lot of memory.
• Analyzing object retention paths (dominator tree) to understand why objects are not collected by the garbage collector.
• Using reports on suspicious leaks, such as "Leak Suspects" in MAT.

How to automate the process:

• Configuring the JVM to automatically dump on memory errors.
• Writing monitoring scripts that trigger jcmd to take a heap dump when a certain memory threshold is reached.
• Integrating with monitoring systems (Prometheus, Grafana) to automatically respond to anomalies and collect dumps.

Example command for manually taking a heap dump:

jcmd <pid> GC.heap_dump /path/to/heapdump.hprof

Bash automation example:

#!/bin/bash
THRESHOLD=80 # % memory usage
PID=$(pidof java_app)
USAGE=$(jstat -gc $PID | tail -1 | awk '{print $3+$5}')
if (( $(echo "$USAGE > $THRESHOLD" | bc -l) )); then
  jcmd $PID GC.heap_dump /tmp/heapdump_$(date +%s).hprof
fi