Testing

Testing#

This is an overview documentation for testing SmartSim and SmartRedis

SmartSim#

SmartSim utilizes pytest for running its test suite. In the top level of SmartSim, users can run multiple testing commands with the developer Makefile.

To run, execute make plus one of the below commands from the top level of the SmartSim directory:

Test
-------
test                - Build and run all tests
test-verbose        - Build and run all tests [verbosely]
test-debug          - Build and run all tests with debug output
test-cov            - Run tests with coverage
test-full           - Run all WLM tests with Python coverage (full test suite)
                      WARNING: do not run test-full on shared systems.

Note

You must have the extra dev dependencies installed in your python environment to execute tests. Install dev dependencies with pip install -e .[dev]

Test Suites#

There are three test suite levels within SmartSim

  • local

  • on_wlm

  • full

local#

local runs by default and doesn’t launch any jobs out onto a system through a workload manager like Slurm. All jobs are contained on the local machine.

This is the test suite that runs in GitHub actions each time a commit is used.

To run the local tests

bash
make test
# or
make test-cov # for coverage
# or
make test-debug # for CLi logging output

on_wlm#

This is the same test suite as the local test suite with the addition of the tests located within the on_wlm directory.

To run the on_wlm test suite, users will have to be on a system with one of the supported workload managers. Additionally, users will need to obtain an allocation of at least 8 nodes.

Examples of how to obtain allocations on systems with the launchers:

# for slurm (with srun)
salloc -N 8 -A account --exclusive -t 00:10:00

# for PBSPro (with aprun)
qsub -l select=8 -l place=scatter -l walltime=00:10:00 -q queue

# for LSF (with jsrun)
bsub -Is -W 00:30 -nnodes 8 -P project $SHELL

Values for queue, account, or project should be substituted appropriately.

Once in an iterative allocation, users will need to set the test launcher environment variable: SMARTSIM_TEST_LAUNCHER to one of the following values

  • slurm

  • pbs

  • lsf

  • local

In addition to the SMARTSIM_TEST_LAUNCHER variable, there are a few other runtime test configuration options for SmartSim

  • SMARTSIM_TEST_LAUNCHER: Workload manager of the system (local by default)

  • SMARTSIM_TEST_ACCOUNT: Project account for allocations (used for customer systems mostly)

  • SMARTSIM_TEST_DEVICE: cpu or gpu

  • SMARTSIM_TEST_NUM_GPUS: the number of GPUs to use for model and script testing (defaults to 1)

  • SMARTSIM_TEST_PORT: the port to use for database communication (defaults to 6780)

  • SMARTSIM_TEST_INTERFACE: network interface to use.

For the SMARTSIM_TEST_INTERFACE, the default is ipogif0 which is the high speed network on Horizon, and other XC systems with the Aries interconnect.

Other possible values are:

  • ipogif0

  • ib0 (and other ib variants)

  • eth0

For the local test suite, the network interface does not need to be set.

A full example on an internal SLURM system

salloc -N 8 -A account --exclusive -t 03:00:00
export SMARTSIM_TEST_LAUNCHER=slurm
export SMARTSIM_TEST_INTERFACE=ipogif0
export SMARTSIM_TEST_DEVICE=gpu
make test-debug

full_wlm#

The full test suite runs the on_wlm tests in addition to tests that will allocate and run on their own allocations. This is the only way that the batch interface is tested.

Unless you know what you’re doing, do not run this on customer systems

Writing Tests for SmartSim

When you introduce new code, it’s imperative that tests accompany your PR. Below are some guidelines for writing new tests.

  • All test files start with test_

  • All test functions start with test_

  • Function name should signal whats being tested

  • All static test files should go in SmartSim/tests/test_configs

  • All test output should be located in SmartSim/tests/test_output (see below on conftest.py)

Write most tests within the base SmartSim/tests directory unless they are meant to specifically test a launcher integration that necessitates its placement into the on_wlm or the full_wlm directory.

Any tests that run AI/ML tests for the backend should be placed in the SmartSim/tests/backend directory.

Most tests utilize the convenience functions with conftest.py which can be added as arguments to test functions to supply runtime values for test directories and what not. See tests for examples.

SmartRedis#

To build and run all tests, run the following command in the top level of the smartredis repository.

make test

Note

The tests require:

  • GCC >= 5

  • CMake >= 3.13

Since these are usually system libraries, we do not install them for the user.

make test-c         # run C tests
make test-fortran   # run Fortran tests. Implicitly, SR_FORTRAN=ON
make test-cpp       # run all C++ tests
make unit-test-cpp  # run unit tests for C++
make test-py        # run Python tests. Implicitly, SR_PYTHON=ON
make testpy-cov     # run python tests with coverage. Implicitly, SR_PYTHON=ON SR_BUILD=COVERAGE
make testcpp-cpv    # run cpp unit tests with coverage. Implicitly, SR_BUILD=COVERAGE

Customizing the Tests#

Several Make variables can adjust the manner in which tests are run:

  • SR_BUILD: change the way that the SmartRedis library is built. (supported: Release, Debug, Coverage; default for testing is Debug)

  • SR_FORTRAN: enable Fortran language build and testing (default is OFF)

  • SR_PYTHON: enable Python language build and testing (default is OFF)

  • SR_TEST_PORT: change the base port for Redis communication (default is 6379)

  • SR_TEST_NODES: change the number of Redis shards used for testing (default is 3)

  • SR_TEST_REDIS_MODE: change the type(s) of Redis servers used for testing. Supported is Clustered, Standalone, UDS; default is Clustered)

  • SR_TEST_REDISAI_VER: change the version of RedisAI used for testing. (Default is v1.2.3; the parameter corresponds the the RedisAI gitHub branch name for the release)

  • SR_TEST_DEVICE: change the type of device to test against. (Supported is cpu, gpu; default is cpu)

  • SR_TEST_PYTEST_FLAGS: tweak flags sent to pytest when executing tests (default is -vv -s)

These variables are all orthogonal. For example, to run tests for all languages against a standalone Redis server, execute the following command:

make test SR_FORTRAN=ON SR_PYTHON=ON SR_TEST_REDIS_MODE=Standalone

Similarly, it is possible to run the tests against each type of Redis server in sequence (all tests against a standalone Redis server, then all tests against a Clustered server, then all tests against a standalone server with a Unix domain socket connection) via the following command:

make test SR_FORTRAN=ON SR_PYTHON=ON SR_TEST_REDIS_MODE=All

Note

Unix domain socket connections are not supported on MacOS. If the SmartRedis test system detects that it is running on MacOS, it will automatically skip UDS testing.

Writing Tests for SmartRedis#

Below are some guidelines for writing new tests. These are fairly similar to SmartSim.

  • Place tests for each client in their language directory (i.e. tests for C client in SmartRedis/tests/c)

  • All test files start with test_

  • All test functions start with test_

  • Function name should signal what’s being tested

Writing Integration Tests#

The integrations tests are run with the pytest framework and some helper python files that spin up the client drivers. Follow the naming convention above and the tests will be included.

Writing Unit Tests#

All unit tests for the C++ client are located at tests/cpp/unit-tests/ and use the Catch2 test framework. The unit tests mostly follow a Behavior Driven Development (BDD) style by using Catch2’s SCENARIO, GIVEN, WHEN, and THEN syntax.

Files that contain Catch2 unit tests should be prefixed with test_ in order to keep a consistent naming convention.

When adding new unit tests, create a new SCENARIO in the appropriate file. If no such file exists, then it is preferred that a new file (prefixed with test_) is created.

  • New unit tests should be placed in tests/cpp/unit-tests/

  • Testing files should be prefixed with test_

  • It is preferred that new unit tests are in a new SCENARIO

Continuous Integration (CI)#

GitHub Actions is our public facing CI that runs in the GitHub cloud.

The actions are defined using yaml files are are located in the .github/workflows/ directory of SmartSim and SmartRedis.

Each pull request, push and merge the test suite for SmartRedis and SmartSim are run. For SmartSim, this is the local test suite with the local launcher.

Contents