Awesome

ULP OBSERVABILITY PLUGIN

UbikLoadPack observability plugin is an extension for Apache JMeter which allows you to monitor a performance test (Standalone or distributed) from your browser.

It also exposes an Open Metrics endpoint so that Promtheus or similar tools can scrape it and make metrics of your performance test available in it.

It is easily installable through JMeter-Plugins manager

Connect the plugin to Prometheus

### Install Prometheus If you haven't installed prometheus on your device yet, Download the latest release of Prometheus for your platform, then extract it. If you use Windows, add the extracted directory to the PATH env variable. For Linux users, add it to the $PATH variable. Add it to the .bashrc file to make it permanent.

Prometheus configuration

Inside the directory of your JMeter test plan, add the following prometheus.yml file :

global:
  scrape_interval:     15s # By default, scrape targets every 15 seconds.

  # Attach these labels to any time series or alerts when communicating with
  # external systems (federation, remote storage, Alertmanager).
  external_labels:
    monitor: 'codelab-monitor'

# A scrape configuration containing Observability plugin endpoint to scrape:
scrape_configs:
  - job_name: 'observability-plugin'
    scrape_interval: 5s
    metrics_path: /ulp-o-metrics

    static_configs:
      - targets: ['localhost:9090']

Change 'localhost:9090' to match the port you are using to host the plugin.

### Prometheus launch Cd to the test plan directory, then enter this command to start using prometheus (on port 9095, change it with the one you prefer) :

prometheus  --web.enable-admin-api   --web.listen-address=:9095 --config.file=prometheus.yml

You can then access Prometheus from this url : http://localhost:9095/graph

Metrics to display:

• Name of the Sampler used for the load test
• Number of requests
• % Error
• Average time
• Percentiles 1 (aggregate_rpt_pct1 property)
• Percentiles 2 (aggregate_rpt_pct2 property)
• Percentiles 3 (aggregate_rpt_pct3 property)
• Max time
• Throughput in req/s

Constraints:

• Have the lightest possible memory / CPU footprint on JMeter performances
• Ability to handle up to 1 million requests per minute
• Select a Javascript graphing library that is free and Open Source compatible or paid but in One Shot only
• Develop the solution in Plugin
• Must not block the caller
• Use a light server in terms of memory / CPU consumption (Embedded Jetty)
• Java 11 compatible code

Technologies and dependencies:

Back

• Java : Language used by JMeter
• Embedded Jetty : to expose the metrics collected by SamplerListener (specific to JMeter) and calculated by HDRHistogram offers a very light server implementation
• Micrometer : to save samples and calculate more metrics
• CronScheduler : for a clock drift insensitive task scheduler
• openapi-generator-maven-plugin : to generate APIs and models
• Jackson : for API response serialization
• Junit : Unit tests

Completed tasks

• Multithreaded sample processing
• Distributed testing support
• Custom sampler configuration
• Metrics logging
• Metrics available in OpenMetrics format through the Jetty server
• HTML page with metric charts
• Javadoc

Screenshots

Front

• TypeScript : to take advantage of a strongly typed language for more rigor and data consistency
• AngularJs: for the front and graphing part : - States and hooks to facilitate data manipulation (Metrics) - npm repo for various dependencies - ChartJs available - Material UI for elegant rendering
• ChartJs : Library used a lot, maintained (last update on 02/16/2022) and free

Completed tasks

• Synchronization with plugin configuration
• HTML page with metrics correctly displays multi-axis charts for each type of metric
• Summary of total metrics at the bottom of the page

Screenshots

• Left y-axis: metric of each group of samples
• Right y-axis: cumulative number of threads in each group of samples
• Graphs currently implemented for: average response, maximum response, percentiles, error percentage and throughput

Roadmap