hedra 0.7.13

Performance testing at scale.

Hedra - Testing at scale

Package	Hedra
Version	0.7.13
Web	https://hedra.dev
Download	https://pypi.org/project/hedra/
Source	https://github.com/scorbettUM/hedra
Keywords	performance, testing, async, distributed, graph, DAG, workflow

Hedra is a Python performance and scalable unit/integration testing framework that makes creating and running complex test workflows easy.

These workflows are written as directed acrylic graphs in Python, where each graph is specified as a collection of Python classes referred to as stages. Each Stage may then specify async Python methods which are then wrapped in Python decorators (referred to as hooks), which that Stage will then execute. The hook wrapping a method tells Hedra both what the action does and when to execute it. In combination, stages and hooks allow you to craft test workflows that can mimic real-world user behavior, optimize framework performance, or interact with a variety of Hedra's powerful integrations.

---

Why Hedra?

Understanding how your application performs under load can provide valuable insights - allowing you to spot issues with latency, memory usage, and stability. However, performance test tools providing these insights are often difficult to use and lack the ability to simulate complex user interaction at scale. Hedra was built to solve these problems by allowing developers and test engineers to author performance tests as sophisticated and scalable workflows. Hedra adheres to the following tenants:

Speed and efficiency by default

Regardless of whether running on your personal laptop or distributed across a cluster, Hedra is fast, capable of generating millions of requests or interactions per minute and without consuming excessive memory. Hedra pushes the limits of Python to achieve this, embracing the latest in Python async and multiprocessing language features to achieve optimal execution performance.

Run with ease anywhere

Authoring, managing, and running test workflows is easy. Hedra includes integrations with Git to facilitate easy management of collections of graphs via Projects, the ability to generate flexible starter test templates, and an API that is both fast and intuitive to understand. Distributed use almost exactly mirrors local operation, reducing the learning curve for more complex deployments.

Flexibility and and painless extensibility

Hedra ships with support for HTTP, HTTP2, Websockets, and UDP out of the box. GraphQL, GRPC, and Playwright are available simply by installing the (optional) dependency packages. Hedra offers JSON and CSV results output by default, with 28 additional results reporting options readily available by likewise installing the required dependencies.

Likewise, Hedra offers a comprehensive plugin system. You can easily write a plugin to test your Postgresql database or integrate a third party service, with CLI-generated templates to guide you and full type hints support throughout. Unlike other frameworks, no additional compilation or build steps are required - just write your plugin, import it, and include it in the appropriate Stage in your test graph.

---

Requirements and Getting Started

Hedra has been tested on Python versions 3.8.6+, though we recommend using Python 3.10+. You should likewise have the latest LTS version of OpenSSL, build-essential, and other common Unix dependencies installed (if running on a Unix-based OS).

Warning: Hedra has currently only been tested on the latest LTS versions of Ubuntu and Debian. Other official OS support is coming Mar. 2023.

Installing

To install Hedra run:

pip install hedra

Verify the installation was was successful by running the command:

hedra --help

which should output

Creating your first graph

Get started by running Hedra's:

hedra graph create <path/to/graph_name.py>

command in an empty directory to generate a basic test from a template. For example, run:

hedra graph create example.py

which will output the following:

and generate the the test below in the specified example.py file:

from hedra import (
	Setup,
	Execute,
	action,
	Analyze,
	JSONConfig,
	Submit,
	depends,
)

class SetupStage(Setup):
    batch_size=1000
    total_time='1m'


@depends(SetupStage)
class ExecuteHTTPStage(Execute):

    @action()
    async def http_get(self):
        return await self.client.http.get('https://<url_here>')


@depends(ExecuteHTTPStage)
class AnalyzeStage(Analyze):
    pass


@depends(AnalyzeStage)
class SubmitJSONResultsStage(Submit):
    config=JSONConfig(
        events_filepath='./events.json',
        metrics_filepath='./metrics.json'
    )

We'll explain this graph below, but for now - replace the string 'https://<url_here>' with 'https://httpbin.org/get'.

Before running our test, if on a Unix system, we may need to set the maximum number of open files above its current limit. This can be done by running:

ulimit -n 256000

note that you can provide any number here, as long as it is greater than the batch_size specified in the SetupStage Stage. With that, we're ready run our first test by executing:

hedra graph run example.py

Hedra will load the test graph file, parse/validate/setup the stages specified, then begin executing your test:

The test will take a minute or two to run, but once complete you should see:

You have officially created and run your first test graph!

---

Development

Local development requires at-minimum Python 3.8.6, though 3.10.0+ is recommended. To setup your environment run:

python3 -m venv ~/.hedra && \
source ~/.hedra/bin/activate && \
git clone https://github.com/scorbettUM/hedra.git && \
cd hedra && \
pip install --no-cache -r requirements.in && \
python setup.py develop

To develop or work with any of the additional provided engines, references the dependency tables below.

---

Engines, Personas, Algorithms, and Reporters

Much of Hedra's extensibility comes in the form of both extensive integrations/options and plugin capabilities for four main framework features:

Engines

Engines are the underlying protocol or library integrations required for Hedra to performance test your application (for example HTTP, UDP, Playwright). Hedra currently supports the following Engines, with additional install requirements shown if necessary:

Engine	Additional Install Option	Dependencies
HTTP	N/A	N/A
HTTP2	N/A	h2, hpack
UDP	N/A	N/A
Websocket	N/A	N/A
GRPC	pip install hedra[grpc]	grpcio grpco-tools, h2, hpack
GraphQL	pip install hedra[graphql]	gql
GraphQL-HTTP2	pip install hedra[graphql]	gql, h2, hpack
Playwright	pip install hedra[playwright] && playwright install	playwright

Personas

Personas are responsible for scheduling when @action() or @task() hooks execute over the specified Execute stage's test duration. No additional install dependencies are required for Personas, and the following personas are currently supported out-of-box:

Persona	Setup Config Name	Description
Batched	batched	Executes each action or task hook in batches of the specified size, with an optional wait between each batch spawning
Constant Arrival Rate	constant-arrival	Hedra automatically adjusts the batch size after each batch spawns based upon the number of hooks that have completed, attempting to achieve batch_size completions per batch
Constant Spawn Rate	constant-spawn	Like Batched, but cycles through actions before waiting batch_interval time.
Default	N/A	Cycles through all action/task hooks in the Execute stage, resulting in a (mostly) even distribution of execution
No-Wait	no-wait	Cycles through all action/task hooks in the Execute stage, resulting in a (mostly) even distribution of execution. WARNING: This persona may cause OOM as it does not use the memory management algorithm.
Ramped	ramped	Starts at a batch size of batch_gradient * batch_size. Batch size increases by the gradient each batch with an optional wait between each batch spawning
Ramped Interval	ramped-interval	Executes batch_size hooks before waiting batch_gradient * batch_interval time. Interval increases by the gradient each batch
Sorted	sorted	Executes each action/task hook in batches of the specified size and in the order provided to each hook's (optional) order parameter
Weighted	weighted	Executes action/task hooks in batches of the specified size, with each batch being generated from a sampled distribution based upon that action's weight

Algorithms

Algorithms are used by Hedra Optimize stages to calculate maximal test config options like batch_size, batch_gradient, and/or batch_interval. All out-of-box supported algorithms use scikit-learn and include:

Algorithm	Setup Config Name	Description
SHG	shg	Uses scikit-learn's Simple Global Homology (SHGO) global optimization algorithm
Dual Annealing	dual-annealing	Uses scikit-learn's Dual Annealing global optimization algorithm
Differential Evolution	diff-evolution	Uses scikit-learn's Differential Evolution global optimization algorithm

Reporters

Reporters are the integrations Hedra uses for submitting aggregated and unaggregated results (for example, to a MySQL database via the MySQL reporter). Hedra currently supports the following Reporters, with additional install requirements shown if necessary:

Engine	Additional Install Option	Dependencies
AWS Lambda	pip install hedra[aws]	boto3
AWS Timestream	pip install hedra[aws]	boto3
Big Query	pip install hedra[google]	google-cloud-bigquery
Big Table	pip install hedra[google]	google-cloud-bigtable
Cassandra	pip install hedra[cassandra]	cassandra-driver
Cloudwatch	pip install hedra[aws]	boto3
CosmosDB	pip install hedra[azure]	azure-cosmos
CSV	N/A	N/A
Datadog	pip install hedra[datadog]	datadog
DogStatsD	pip install hedra[statsd]	aio_statsd
Google Cloud Storage	pip install hedra[google]	google-cloud-storage
Graphite	pip install hedra[statsd]	aio_statsd
Honeycomb	pip install hedra[honeycomb]	libhoney
InfluxDB	pip install hedra[influxdb]	influxdb_client
JSON	N/A	N/A
Kafka	pip install hedra[kafka]	aiokafka
MongoDB	pip install hedra[mongodb]	motor
MySQL	pip install hedra[sql]	aiomysql, sqlalchemy
NetData	pip install hedra[statsd]	aio_statsd
New Relic	pip install hedra[newrelic]	newrelic
Postgresql	pip install hedra[sql]	aiopg, psycopg2-binary, sqlalchemy
Prometheus	pip install hedra[prometheus]	prometheus-client, prometheus-client-api
Redis	pip install hedra[redis]	redis, aioredis
S3	pip install hedra[aws]	boto3
Snowflake	pip install hedra[snowflake]	snowflake-connector-python, sqlalchemy
SQLite3	pip install hedra[sql]	sqlalchemy
StatsD	pip install hedra[statsd]	aio_statsd
Telegraf	pip install hedra[statsd]	aio_statsd
TelegrafStatsD	pip install hedra[statsd]	aio_statsd
TimescaleDB	pip install hedra[sql]	aiopg, psycopg2-binary, sqlalchemy

---

Resources

Hedra's official and full documentation is currently being written and will be linked here soon!

---

License

This software is licensed under the MIT License. See the LICENSE file in the top distribution directory for the full license text.

---

Contributing

Hedra will be open to general contributions starting Fall, 2023 (once the distributed rewrite and general testing is complete). Until then, feel free to use Hedra on your local machine and report any bugs or issues you find!

---

Code of Conduct

Hedra has adopted and follows the Contributor Covenant code of conduct. If you observe behavior that violates those rules please report to:

Name	Email	Twitter
Sean Corbett	[e-mail]	@sc_codeum