Master JVM Performance: Essential Tools and Real‑World Debugging Guide

1. jps – Basic JVM Process Status Tool

jps lists the Java processes running on a host. Syntax: jps [options] [hostid]. Common options are -q (no class/JAR name), -m (show main arguments), -l (full package name), and -v (JVM arguments). Example output shows process IDs, main classes, and arguments.

2. jstack – Thread Stack Dump

jstack prints stack traces of a Java process. Syntax: jstack [option] pid (or for core files and remote hosts). The -l option adds lock information, useful for dead‑lock analysis. A typical workflow to find the most CPU‑intensive thread is:

Identify the Java process ID with ps -ef | grep mrf-center.

Find the thread with highest CPU usage using top -Hp <pid> and note its numeric ID (e.g., 21742).

Convert the thread ID to hexadecimal (e.g., 0x54ee).

Run jstack <pid> | grep 54ee to locate the corresponding Java thread and stack trace.

The example reveals a thread named PollIntervalRetrySchedulerThread blocked in Object.wait(), and the related source code snippet is shown.

java -agentlib:hprof=cpu=samples,interval=20,depth=3 Hello

3. jmap and jhat – Heap Dump and Analysis

jmap exports heap information; jhat analyzes the dump. Syntax: jmap [option] pid. Useful options include: -heap – shows heap configuration and usage. -permstat – prints permanent generation statistics. -histo[:live] – prints a histogram of object counts and sizes. -dump:format=b,file=<path> – creates a binary heap dump.

Sample jmap -heap 21711 output displays GC algorithm, heap sizes, and per‑generation usage. The histogram command jmap -histo:live 21711 lists the most frequent object types. The binary dump can be opened with MAT, VisualVM, or jhat:

jhat -port 9998 /tmp/dump.dat

4. jstat – JVM Statistics Monitoring

jstat reports runtime statistics such as memory pool sizes and garbage‑collection activity. Syntax: jstat [generalOption|outputOptions] vmid [interval [count]]. Example showing GC metrics every 250 ms for four samples: jstat -gc 21711 250 4 The columns are explained (S0C/S1C – Survivor spaces, EC/EU – Eden, OC/OU – Old generation, PC/PU – Permanent generation, YGC/YGCT – young GC count/time, FGC/FGCT – full GC count/time, GCT – total GC time). An accompanying diagram illustrates the JVM heap layout.

5. hprof – Heap/CPU Profiling Tool

hprof can profile CPU usage and heap allocation. Invocation forms include:

java -agentlib:hprof[=options] Class
java -Xrunprof[:options] Class
javac -J-agentlib:hprof[=options] Class

Key options (default values in brackets) are: heap=dump|sites|all – heap profiling mode. cpu=samples|times|old – CPU profiling mode. monitor=y|n – monitor lock contention. format=a|b – text or binary output. file=<file> – output file name (default java.hprof[.txt]). interval=<ms> – sampling interval (default 10 ms).

… (other options omitted for brevity).

Example commands:

java -agentlib:hprof=cpu=samples,interval=20,depth=3 Hello
javac -J-agentlib:hprof=cpu=times Hello.java
javac -J-agentlib:hprof=heap=sites Hello.java
javac -J-agentlib:hprof=heap=dump Hello.java

Note: using -Xrunprof on production servers incurs significant overhead and is not recommended.