Skip to content

alteryx/locust-grasshopper

Repository files navigation

Grasshopper

A lightweight framework for performing load tests against an environment, primarily against an API. Grasshopper glues Locust, Pytest, some plugins (namely Locust InfluxDBListener ) and some custom code to provide a package that makes authoring load tests simple with very little boilerplate needed.

Here are some key functionalities that this project extends on Locust:

Installation

This package can be installed via pip: pip install locust-grasshopper

Example Load Test

  • You can refer to the test test_example.py in the example directory for a basic skeleton of how to get a load test running. In the same directory, there is also an example conftest.py that will show you how to get basic parameterization working.
  • This test can be invoked by running pytest example/test_example.py in the root of this project.
  • This test can also be invoked via a YAML scenario file:
cd example
pytest example_scenarios.YAML --tags=example1

In this example scenario file, you can see how grasshopper_args, grasshopper_scenario_args, and tags are being set.

(back to top)

Creating a load test

When creating a new load test, the primary grasshopper function you will be using is called Grasshopper.launch_test. This function can be imported like so: from grasshopper.lib.grasshopper import Grasshopper launch_test takes in a wide variety of args:

  • user_classes: User classes that the runner will run. These user classes must extend BaseJourney, which is a grasshopper class (from grasshopper.lib.journeys.base_journey import BaseJourney). This can be a single class, a list of classes, or a dictionary where the key is the class and the value is the locust weight to assign to that class.
  • **complete_configuration: In order for the test to have the correct configuration, you must pass in the kwargs provided by the complete_configuration fixture. See example load test on how to do this properly.

(back to top)

Scenario Args

  • If you want to parameterize your journey class, you should use the scenario_args dict. This is the proper way to pass in values from outside of the journey for access by the journey code. Note that each journey gets a copy on start, so the journey itself can safely modify its own dictionary once the test is running. scenario_args exists for any journey that extends the grasshopper base_journey class. scenario_args also grabs from self.defaults on initialization. For example:
from locust import between, task
from grasshopper.lib.journeys.base_journey import BaseJourney
from grasshopper.lib.grasshopper import Grasshopper

# a journey class with an example task
class ExampleJourney(BaseJourney):
    # number of seconds to wait between each task
    wait_time = between(min_wait=20, max_wait=30)
    
    # this defaults dictionary will be merged into scenario_args with lower precedence 
    # when the journey is initialized
    defaults = {
        "foo": "bar",
    }
    
    @task
    def example_task:
        logging.info(f'foo is `{self.scenario_args.get("foo")}`.')
        
        # aggregate all metrics for the below request under the name "get google"
        # if name is not specified, then the full url will be the name of the metric
        response = self.client.get('https://google.com', name='get google')

# the pytest test which launches the journey class
def test_run_example_journey(complete_configuration):
    # update scenario args before initialization
    ExampleJourney.update_incoming_scenario_args(complete_configuration)
    
    # launch the journey
    locust_env = Grasshopper.launch_test(ExampleJourney, **complete_configuration)
    return locust_env

(back to top)

Commonly used grasshopper pytest arguments

  • --runtime: Number of seconds to run each test. Set to 120 by default.
  • --users: Max number of users that are spawned. Set to 1 by default.
  • --spawn_rate : Number of users to spawn per second. Set to 1 by default.
  • --shape: The name of a shape to run for the test. If you don't specify a shape or shape instance, then the shape Default will be used, which just runs with the users, runtime & spawn_rate specified on the command line (or picks up defaults of 1, 1, 120s). See utils/shapes.py for available shapes and information on how to add your own shapes.
  • --scenario_file If you want a yaml file where you pre-define some args, this is how you specify that file path. For example, scenario_file=example/scenario_example.YAML.
  • --scenario_name If --scenario_file was specified, this is the scenario name that is within that YAML file that corresponds to the scenario you wish to run. Defaults to None.
  • --tags See below example: Loop through a collection of scenarios that match some tag.
  • --scenario_delay Adds a delay in seconds between scenarios. Defaults to 0.
  • --influx_host If you want to report your performance test metrics to some influxdb, you must specify a host. E.g. 1.1.1.1. Defaults to None.
  • --influx_port: Port for your influx_host in the case where it is non-default.
  • --influx_user: Username for your influx_host, if you have one.
  • --influx_pwd: Password for your influx_host, if you have one.
  • --grafana_host: If your grafana is a separate URL from the influxdb, you can specify it here. If you don't, then the grafana URL will be the same as the influxdb URL when the grasshopper object generates grafana links.

(back to top)

Launching tests with a configuration

All in all, there are a few ways you can actually collect and pass params to a test:

Run a test with its defaults

cd example pytest test_example.py ...

Run a test with a specific scenario

cd example pytest test_example.py --scenario_file=example_scenarios.YAML --scenario_name=example_scenario_1 ...

Loop through a collection of scenarios that match some tag

cd example pytest example_scenarios.YAML --tags=smoke ...

  • As shown above, this case involves passing a .YAML scenario file into pytest instead of a .py file.
  • The --scenario_file and --scenario_name args will be ignored in this case
  • The --tags arg supports AND/OR operators according to the opensource tag-matcher package. More info on these operators can be found here.
  • If no --tags arg is specified, then ALL the scenarios in the .yaml file will be run.
  • If a value is given for --scenario_delay, the test will wait that many seconds between collected scenarios.
  • All scenarios are implicitly tagged with the scenario name to support easily selecting one single scenario

(back to top)

Configuring Grasshopper

Grasshopper adds a variety of parameters relating to performance testing along with a variety of ways to set these values.

Recent changes (>= 1.1.1) include an expanded set of sources, almost full access to all arguments via every source (some exceptions outlined below), and the addition of some new values that will be used with integrations such as report portal & slack (integrations are NYI). These changes are made in a backwards compatible manner, meaning all existing grasshopper tests should still run without modification. The original fixtures and sources for configuration are deprecated, but still produce the same behavior.

All of the usual pytest arguments also remain available.

The rest of the sections on configuration assume you are using locust-grasshopper>=1.1.1.

(back to top)

Sources

Currently, there are 5 different sources for configuration values and they are, in precedence order

  • command line arguments
  • environment variables
  • scenario values from a scenario yaml file
  • grasshopper configuration file
  • global defaults (currently stored in code, not configurable by consumers)

Obviously, the global defaults defined by Grasshopper are not really a source for consumers to modify, but we mention it so values don't seem to appear "out of thin air".

(back to top)

Categories

The argument list is getting lengthy, so we've broken it down into categories. These categories are entirely for humans: better readability, understanding and ease of use. Once they are all fully loaded by Grasshopper, they will be stored in a single GHConfiguration object (dict). By definition, every argument is in only one category and there is no overlap of keys between the categories. If the same key is supplied in multiple categories, they will be merged with the precedence order as they appear in the table.

Name Scope Description/Usage
Grasshopper Session Variables that rarely change, may span many test runs.
Test Run Session Variables that may change per test run, but are the
same for every scenario in the run
Scenario Session Variables that may change per scenario and are often
scenario specific; Includes user defined variables that are
not declared as command line arguments by Grasshopper.
However, you may use pytest's addoptions hook in your
conftest to define them.

At least one of the sections must be present in the global configuration file and eventually this will be the same in the configuration section of a scenario in a scenario yaml file. Categories are not used when specifying environment variables or command line options. We recommend that you use these categories in file sources, but if a variable is in the wrong section, it won't actually affect the configuration loading process.

(back to top)

Using Configuration Values

Your test(s) may access the complete merged set of key-value pairs via the session scoped fixture complete_configuration. This returns a GHConfiguration object (dict) which is unique to the current scenario. This value will be re-calculated for each new scenario executed.

A few perhaps not obvious notes about configuration:

  • use the environment variable convention of all uppercase key names (e.g. RUNTIME=10) to specify a key-value pair via an environment value
  • use the lower case key to access a key from the GHConfiguration object (e.g. x = complete_configuration("runtime")) regardless of the original source(s)
  • use -- before the key name to specify it on the command line (e.g. --runtime=10)
  • configure a grasshopper configuration file by creating a session scoped fixture loaded by your conftest.py called grasshopper_config_file_path which returns the full path to a configuration yaml file.
  • grasshopper supports thresholds specified as
    • a json string - required for environment variable or commandline, but also accepted from other sources
    • a dict - when passing in via the scenario_args method (more details on that below) or via a journey class's defaults attr.
@pytest.fixture(scope="session")
def grasshopper_config_file_path():
    return "path/to/your/config/file"

An example grasshopper configuration file:

grasshopper:
  influx_host: 1.1.1.1
test_run:
  users: 1.0
  spawn_rate: 1.0
  runtime: 600
scenario:
  key1 : 'value1'
  key2: 0

(back to top)

Additional Extensions to the configuration loading process

If you would like to include other environment variables that might be present in the system, you can define a fixture called extra_env_var_keys which returns a list of key names to load from the os.environ. Keys that are missing in the environment will not be included in the GHConfiguration object.

Any environment variables that use the prefix GH_ will also be included in the GHConfiguration object. The GH_ will be stripped before adding and any names that become zero length after the strip will be discarded. This is a mechanism to include any scenario arguments you might like to pass via an environment variable.

In the unlikely case that you need to use a different prefix to designate scenario variables, you can define a fixture called env_var_prefix_key which returns a prefix string. The same rules apply about which values are included in the configuration.

(back to top)

Checks

Checks are an assertion that is recorded as a metric. They are useful both to ensure your test is working correctly (e.g. are you getting a valid id back from some post that you sent) and to evaluate if the load is causing intermittent failures (e.g. sometimes a percentage of workflow runs don't complete correctly the load increases). At the end of the test, checks are aggregated by their name across all journeys that ran and then reported to the console. They are also forwarded to the DB in the "checks" table. Here is an example of using a check:

from grasshopper.lib.util.utils import check
...
response = self.client.get(
    'https://google.com', name='get google'
)
check(
    "get google responded with a 200",
    response.status_code == 200,
    env=self.environment,
)

It is worth noting that it is NOT necessary to add checks on the http codes. All the HTTP return codes are tracked automatically by grasshopper and will be sent to the DB. If you aren't using a DB then you might want the checks console output.

(back to top)

Custom Trends

Custom trends are useful when you want to time something that spans multiple HTTP calls. They are reported to the specified database just like any other HTTP request, but with the "CUSTOM" HTTP verb as opposed to "GET", "POST", etc. Here is an example of using a custom trend:

from locust import between, task
from grasshopper.lib.util.utils import custom_trend
...

@task
@custom_trend("my_custom_trend")
def google_get_journey(self)
    for i in range(len(10)):
        response = self.client.get(
            'https://google.com', name='get google', context={'foo1':'bar1'}
        )

(back to top)

Thresholds

Thresholds are time-based, and can be added to any trend, whether it be a custom trend or a request response time. Thresholds default to the 0.9 percentile of timings. Here is an example of using a threshold:

# a journey class with an example threshold
from locust import between, task
from grasshopper.lib.journeys.base_journey import BaseJourney
from grasshopper.lib.grasshopper import Grasshopper

class ExampleJourney(BaseJourney):
    # number of seconds to wait between each task
    wait_time = between(min_wait=20, max_wait=30)
    
    @task
    def example_task:
        self.client.get("https://google.com", name="get google")
        
    @task
    @custom_trend("my custom trend")
    def example_task_custom_trend:
        time.sleep(10)

# the pytest test which launches the journey class, thresholds could be 
# parameterized here as well.
def test_run_example_journey(complete_configuration):
    ExampleJourney.update_incoming_scenario_args(complete_configuration)
    ExampleJourney.update_incoming_scenario_args({
        "thresholds": {
            "get google":
                {
                    "type": "get",
                    "limit": 4000  # 4 second HTTP response threshold
                },
            "my custom trend":
                {
                    "type": "custom",
                    "limit": 11000  # 11 second custom trend threshold
                }
        }
    })
    
    locust_env = Grasshopper.launch_test(ExampleJourney, **complete_configuration)
    return locust_env

Thresholds can also be defined for individual YAML scenarios. Refer to the thresholds key in example/example_scenarios.YAML for how to use thresholds for YAML scenarios.

After a test has concluded, trend/threshold data can be found in locust_env.stats.trends. This data is also reported to the console at the end of each test.

(back to top)

Time Series DB Reporting and Tagging

Additional design details about how a database listener works with grasshopper/locust can be found in the Database Listener Design Documentation.

When you specify a time series database URL param to launch_test, such as influx_host, all metrics will be automatically reported to tables within the locust timeseries database via the specified URL. These tables include:

  • locust_checks: check name, check passed, etc.
  • locust_events: test started, test stopped, etc.
  • locust_exceptions: error messages
  • locust_requests: HTTP requests and custom trends

To run the influxdb/grafana locally, you can use the docker-compose file in the example directory:

cd example/observability_infrastructure
docker-compose up -d

and then you can access the grafana UI at localhost. The default username/password is admin/admin. To then run a test which reports to this influxdb just add the --influx_host=localhost handle.

There are a few ways you can pass in extra tags which will be reported to the time series DB:

  1. HTTP Request Tagging
    All HTTP requests are automatically tagged with their name. If you want to pass in extra tags for a particular HTTP request, you can pass them in as a dictionary for the context param when making a request. For example:

    self.client.get('https://google.com', name='get google', context={'foo':'bar'})

    The tags on this metric would then be: {'name': 'get google', 'foo': 'bar'} which would get forwarded to the database if specified.

  2. Check Tagging
    When defining a check, you can pass in extra tags with the tags parameter:

    from grasshopper.lib.util.utils import check
    ...
    response = self.client.get(
    'https://google.com', name='get google', context={'foo1':'bar1'}
    )
    check(
       "get google responded with a 200",
       response.status_code == 200,
       env=self.environment,
       tags = {'foo2': 'bar2'}
    )
  3. Custom Trend Tagging
    Since custom trends are decorators, they do not have access to non-static class variables when defined. Therefore, you must use the extra_tag_keys param, which is an array of keys that exist in the journey's scenario_args. So for example, if a journey had the scenario args {"foo" : "bar"} and you wanted to tag a custom trend based on the "foo" scenario arg key, you would do something like this:

    from locust import between, task
    from grasshopper.lib.util.utils import custom_trend
    ...
        
    @task
    @custom_trend("my_custom_trend", extra_tag_keys=["foo"])
    def google_get_journey(self)
       for i in range(len(10)):
          response = self.client.get(
             'https://google.com', name='get google', context={'foo1':'bar1'}
          )

(back to top)

Project Roadmap

  • Custom Trends
  • Checks
  • Thresholds
  • Tagging
  • InfluxDB metric reporting
  • docker-compose template for influxdb/grafana
  • PrometheusDB metric reporting
  • Slack reporting
  • ReportPortal reporting

See the open issues for a full list of proposed features (and known issues).

(back to top)

Contributing

Contributions are what make the open source community such an amazing place to learn, inspire, and create. Any contributions you make are greatly appreciated.

If you have a suggestion that would make this better, please fork the repo and create a pull request. You can also simply open an issue with the tag "enhancement". Don't forget to give the project a star! Thanks again!

  1. Fork the Project
  2. Create your Feature Branch (git checkout -b feature/AmazingFeature)
  3. Make sure unit tests pass (pytest tests/unit)
  4. Add unit tests to keep coverage up, if necessary
  5. Commit your Changes (git commit -m 'Add some AmazingFeature')
  6. Push to the Branch (git push origin feature/AmazingFeature)
  7. Open a Pull Request

(back to top)