Resource Monitoring Service

A frequent task for tests is to monitor target server's health. Monitoring service is built to collect data from those remote servers. At this time the following sources are supported:

  • local health stats, enabled by default
  • ServerAgent — technology that is used by JMeter users for long time and
  • Graphite

Local Monitoring Stats

This service collects local health stats from computer running Taurus. It is enabled by default.

Following metrics are collected locally:

  • cpu - total CPU usage %
  • mem - total RAM usage %
  • bytes-sent/bytes-recv - network transfer rate
  • disk-read/disk-write - disk I/O rate
  • disk-space - % disk space used for artifacts storage
  • engine-loop - Taurus "check loop" utilization, values higher than 1.0 means you should increase settings.check-interval
  • conn-all - quantity of network connections
- module: monitoring
  - interval: 20s   # polling interval
    - cpu
    - disk-space
    - engine-loop

Once you have resource monitoring enabled, you'll be presented with small sidebar widget that informs you on latest data from monitoring agents:

The widget will possibly not display all the metrics for the long list, that's the limitation of screen height :)


ServerAgent is a small Java application that collects server health stats and makes them accessible through network connection. To use it, you need to install and launch ServerAgent on each of your target servers and then specify metrics to collect under services item. For example:

- module: monitoring
  - address:
    label: target-server  # if you specify label, it will be used in reports instead of ip:port
    interval: 3s    # polling interval
    - cpu
    - disks
    - memory


Graphite data source uses graphite The Render URL API to receive metrics. In this example you can see usage optional server label, timeout for graphite answers, interval between requests and interesting graphite data range definition with parameters from/until.

- module: monitoring
  - address:
    interval: 5s
    from: 100s
    until: 1s
    timeout: 2s
    - store.memUsage
    - test.param1
  - address: local_serv:2222
    label: test_serv
    - production.hardware.cpuUsage
    - groupByNode(myserv_comp_org.cpu.?.cpu.*.value, 4, 'avg')