rally/doc/source/plugins.rst

Scenarios Plugins

Rally provides an opportunity to create and use a custom benchmark scenario as a plugin. The plugins mechanism can be used to simplify some experiments with new scenarios and to facilitate their creation by users who don't want to edit the actual Rally code.

Placement

Put the plugin into the /etc/rally/plugins/scenarios or ~/.rally/plugins/scenarios directory and it will be autoloaded (they are not created automatically, you should create them by hand). The corresponding module should have ".py" extension.

Creation

Inherit a class containing the scenario method(s) from rally.benchmark.scenarios.base.Scenario or its subclasses. Place every atomic action in separate function and wrap it with decorator atomic_action_timer from rally.benchmark.scenarios.utils. Pass action name as a string argument to decorator. This name should be unique for every atomic action. It also will be used to show and store results. Combine atomic actions into your benchmark method and wrap it with the scenario decorator from rally.benchmark.scenarios.base.

Sample

You can run this sample to test whether the plugin has been loaded and benchmark scenario results have been stored correctly.

import random
import time

from rally.benchmark.scenarios import base
from rally.benchmark.scenarios import utils as scenario_utils


class PluginClass(base.Scenario):

    @scenario_utils.atomic_action_timer("test1")
    def _test1(self, factor):
        time.sleep(random.random() * factor)

    @scenario_utils.atomic_action_timer("test2")
    def _test2(self, factor):
        time.sleep(random.random() * factor * 10)

    @base.scenario()
    def testplugin(self, factor=1):
        self._test1(factor)
        self._test2(factor)

Usage

Specify the class and the benchmark method of your plugin at the top level of the benchmark task configuration file. If you need to pass some arguments to the benchmark method, place it in the args section of the task configuration file.

Sample

{
    "PluginClass.testplugin": [
        {
            "args": {
                "factor": 2
            },
            "runner": {
                "type": "constant",
                "times": 3,
                "concurrency": 1
            },
            "context": {
                "users": {
                    "tenants": 1,
                    "users_per_tenant": 1
                }
            }
        }
    ]
}