/jmeter-aws

Config and setup of JMeter workers on AWS

Primary LanguageShell

Distributed JMeter test environment

Quick Start (to connect to existing JMeter workers):

  • Check out this repository.
  • Run ./setup-local to download and extract JMeter.
  • Copy your AWS private key to the certs directory or use ssh-add to add your key to the agent. This is the key you use to log in to the AWS instances.
  • Update config/remotes with your AWS worker node public ip addresses. Format is . I normally use aliases jmeter1, jmeter2, etc.
  • Modify any local settings in config/settings.
  • Run ./update-properties to sync up the config files with remotes.
  • To verify you can connect to a JMeter node run ./connect jmeter1 to open an ssh connection. Then exit.
  • Run ./stop-nodes to stop any running JMeter nodes.
  • Run ./start-nodes to open the SSH tunnels and start the nodes for your test.
  • Run ./start-gui to launch the default JMeter project.

Optional Environment Overrides

  • CONFIG_PATH - Directory of modified config settings. This is used to manage site-specific configuration independent of the jmeter-aws project.
  • DATA_PATH - Directory that contains data files to be used for jmeter tests. All files in this directory will be copied to the worker nodes so they are available locally.

Instructions

  1. Run ./setup-local from this directory This script downloads JMeter 2.13 and unpacks it in the test directory.

  2. Create stock AWS Ubuntu instances for JMeter worker nodes. The t2.micro instances will allow you to run up to 250 threads per node. The jmeter.propertes file may need to be updated to tune for more threads/nodes. Ubuntu instances should be configured with a public IP address and SSH access to the test master. You should use the same .pem key for each instance so we can connect to the machines later.

  3. Record the IP Addresses and a worker node alias in the remotes file. Node aliases do not need to be machine names and do not need to appear in the /etc/hosts file. They are only used as machine shortcuts for the launch scripts.

  4. After making any changes to the remotes file

    • Run ./update-properties This will update the jmeter.properties and settings file with the updated worker node addresses.

  5. Copy the .pem key for the AWS instances to the certs directory. Make sure key is chmod 400 for ssh to connect.

  6. Run ./setup-nodes from this directory This script will connect to each worker node in the remotes file, download meter 2.13, update and install openjdk-7-jdk via apt, and unpack JMeter into the home directory.

  7. To start the worker nodes (for each node):

    • Open a terminal
    • Run ./launch

  8. To start the test client:

    • Open a terminal
    • Run ./start-gui

  9. From the GUI menu: Run -> Remote Start -> Starting nodes seems to work best if you start them one at a time. Otherwise there may be socket contention on the client machine. If you start seeing Too many open files errors in the worker node consoles, try starting one client, let the threads initialize, then the next… and so on.

  10. Collect data and stop and Run -> Remote Stop All Data should be collected in the results folder: $TEST_HOME/results/summary-results.jtl

Other Tools

  • connect - SSH to the worker node alias.
  • connect [command] - Execute a command on the worker node.
  • reboot-nodes - Reboot all of the worker node instances.
  • start-tunnels - Launch SSH reverse tunnels to report back to test master. (requires killall ssh to stop the tunnels)
  • start-nodes - Batch script to start all of the worker nodes and intiate SSH tunnels. (requires killall ssh to stop the tunnels)
  • stop-nodes - Stop the java process running on all worker nodes.
  • start-test - Run the JMeter test headless.