testing.asciidoc

Testing

Beats has a various sets of tests. This guide should help to understand how the different test suites work, how they are used and new tests are added.

In general there are two major test suites:

• Tests written in Go

• Tests written in Python

The tests written in Go use the Go Testing package. The tests written in Python depend on nosetests and require a compiled and executable binary from the Go code. The python test run a beat with a specific config and params and either check if the output is as expected or if the correct things show up in the logs.

For both of the above test suites so called integration tests exists. Integration tests in Beats are tests which require an external system like Elasticsearch to test if the integration with this service works as expected. Beats provides in its testsuite docker containers and docker-compose files to start these environments but a developer can run the required services also locally.

Running Go Tests

The Go tests can be executed in each Go package by running go test .. This will execute all tests which don’t don’t require an external service to be running. To also run the Go integration tests run go test -tags=integration .. It will require you to run the expected services on localhost.

To run all non integration tests for a beat run make unit. This will execute all the tests which are not inside a vendor directory. If you want to have a coverage report for the tests which were run use make unit-tests. A coverage report will be generated under build/coverage directory.

All Go tests are in the same package as the tested code itself and have the postfix _test in the file name. Most of the tests are in the same package as the rest of the code. Some of the tests which should be separate from the rest of the code or should not use private variables go under {packagename}_test.

Running Python Tests

The system tests require a testing binary to be available and the python environment to be set up. To create the testing binary run make {beatname}.test. This will create the test binary in the beat directory. To setup the testing environment make python-env can be run which will use venv to load the dependencies. Then nosetests has to be run inside tests/system.

To automate all these steps into one make system-tests can be run. This creates the binary, the environment and runs all tests which do not require and external service.

For the tests which require an external service like Elasticsearch to be running use INTEGRATION_TESTS=1 make system-tests. This will assume the services are running on localhost.

To run the tests without the services running locally, the command make system-tests-environment can be used. This will start up the required environment in docker and will run all tests including the testing binary inside the docker environment.

All Python tests are under tests/system directory.

Test commands

This is a quick summary of the available test commands:

• unit: Go tests

• unit-tests: Go tests with coverage reports

• integration-tests: Go tests with services in local docker

• integration-tests-environment: Go tests inside docker with service in docker

• fast-system-tests: Python tests

• system-tests: Python tests with coverage report

• INTEGRATION_TESTS=1 system-tests: Python tests with local services

• system-tests-environment: Python tests inside docker with service in docker

• testsuite: Complete test suite in docker environment is run

• test: Runs testsuite without environment

There are two experimental test commands:

• benchmark-tests: Running Go tests with -bench flag

• load-tests: Running system tests with LOAD_TESTS=1 flag

Coverage report

If the tests were run to create a test coverage, the coverage report files can be found under build/docs. To create a more human readable file out of the .cov file make coverage-report can be used. It creates a .html file for each report and a full.html as summary of all reports together in the directory build/coverage.

Race detection

All tests can be run with the Go race detector enabled by setting the environment variable RACE_DETECTOR=1. This applies to tests in Go and Python. For Python the test binary has to be recompile when the flag is changed. Having the race detection enabled will slow down the tests.

Docker environment

Running the tests inside the docker environment is useful to not have to setup all services locally. It has the disadvantage that also the binary is run inside the docker container and not on the local machine.