A thread dump is a list of all the Java threads that are currently active in a Java Virtual Machine (JVM). There  are several ways to take thread dumps from a JVM. It is highly recommended to take more than 1 thread dump while analyzing any problem such as deadlock or resource usage analysis. It is always better to confirm in more than one thread dump then making conclusions in single attempt.

Step 1) Get the PID of your java process

The first piece of information you will need to be able to obtain a thread dump is your java process’s PID.

The java JDK ships with the jps command which lists all java process ids. You can run this command like this:



jps -l



Remember, you may have to run this command as sudo -u jps -l, where “user” is the username of the user that the java process is running as.
Even now, if you are not able to find out process id, use below command:



ps -aef | grep java



Step 2) Request a Thread Dump from the JVM

If installed/available, we recommend using the jstack tool. It prints thread dumps to the command line console.

To obtain a thread dump using jstack, run the following command:



jstack



You can output consecutive thread dumps to a file by using the console output redirect / append directive:



jstack >> threaddumps.log



Important points

  1. The jstack tool is available since JDK 1.5.
  2. jstack works even if the -Xrs jvm parameter is enabled.
  3. It’s not possible to use the jstack tool from JDK 1.6 to take threaddumps from a process running on JDK 1.5.

Thread dump sampling in fixed time intervals using jstack script

This simple shell script takes several jstack snapshots in fixed time intervals: [Reference document]



#!/bin/bash

if [ $# -eq 0 ]; then
    echo >&2 "Usage: jstackSeries  [  [  ] ]"
    echo >&2 "    Defaults: count = 10, delay = 1 (seconds)"
    exit 1
fi

pid=$1          # required
count=${2:-10}  # defaults to 10 times
delay=${3:-1} # defaults to 1 second

while [ $count -gt 0 ]
do
    jstack $pid >jstack.$pid.$(date +%H%M%S.%N)
    sleep $delay
    let count--
    echo -n "."
done



Use above script like this:



jstackSeries 10 5



How to compare the thread dumps

To compare the results you may use interactive diff viewers, e.g.



vimdiff file1 file2 file3 file4 # up to 4 files



Another way to see what parts of the jstack trace are changing over time is to compare adjacent jstack trace using context diff (-c option):



d_old=""
for d in jstack.13585.12171*
do
  if [ -n "$d_old" ]
  then
    diff -c "$d_old" "$d"
  fi
  d_old="$d"
done



Here, the result shows only the places where the jstack trace changes from file to file.