The Matrix native step enables your pipeline to repeatedly execute the same set of actions in a variety of configurations and runtime environments, with each variant executing as an independent “steplet.” These steplets can, when configured, execute in parallel on multiple build nodes. On completion of all steplets, Pipelines aggregates the result status, giving the appearance of a single step.
Common use cases for Matrix steps are:
Splitting a large test suite into smaller units, and executing them in parallel to reduce total execution time.
Testing against multiple values of environment variables or multiple runtime images.
Single Dimension Matrix
The Matrix step is useful for performing the same operation multiple times, but with different settings each time. This could mean pushing the same Docker image to different registries, or executing the same set of unit tests with different test values.
For example, we might want to perform the same test on a completed build with different sets of values for animal, mineral, and vegetable.
animal | mineral | vegetable | |
---|---|---|---|
Variation 1 | dog | copper | carrot |
Variation 2 | goat | iron | broccoli |
Variation 3 | lizard | lead | <null> |
This example pipeline triggers on any change to the BuildInfo resource produced by another pipeline. The Matrix step performs the same execution
actions in parallel steplets, with each using a different set of animal, mineral, and vegetable values from the collection specified in the stepletMultipliers:environmentVariables
block.
- name: Matrix_Single_Pipeline steps: - name: matrix_example type: Matrix stepMode: Bash configuration: inputResources: - name: CompletedBuildInfo stepletMultipliers: environmentVariables: # Sets of environment variables for steplets - animal: dog # - Set 1 mineral: copper vegetable: carrot - animal: goat # - Set 2 mineral: iron vegetable: broccoli - animal: lizard # - Set 3 mineral: lead execution: onExecute: - echo "Executing matrix step on ${steplet_id}" - echo "Variation animal = ${animal}, mineral = ${mineral}, vegetable = ${vegetable}"
When successfully executed, the run log of the pipeline shows the set of steplets executed by the Matrix step, along with their individual success status and the set of values used for the Matrix Combination.
Two-Dimension Matrix
A Matrix step can also perform the same operations in multiple runtime environments. For example, running tests in Java and Javascript, or in different release versions of the same runtime.
This can be done as a single-dimension matrix (performing a single, fixed set of actions in each runtime) or as a two-dimension matrix (performing different actions with different configurations in multiple runtimes).
If we modify the single matrix example above to run in both Java and Javascript runtimes, it will execute in a total of 6 steplets; in 2 runtimes for each of 3 sets of environment variables:
Runtime | Set 1 | Set 2 | Set 3 |
---|---|---|---|
java 13.0 | dog, copper, carrot | goat, iron, broccoli | lizard, lead, <null> |
node 8.17.0 | dog, copper, carrot | goat, iron, broccoli | lizard, lead, <null> |
Matrix Pipeline Example
The following example pipeline can be found in the JFrog GitHub repository. You can fork this repo to your own GitHub account to try on your own installation of Pipelines.
The two-dimension matrix pipeline provides a simplified example of running tests in modules in different runtime environments. It performs basic functionality tests on four different math operation modules, repeating these tests in different release versions of nodejs.
Resources
The first step of the pipeline will trigger on any change to the example GitHub repo, which also houses the test files. So any change to the tests or the pipeline will trigger a new run of the pipeline.
If you have forked the example, you will need to change the path and integration name as noted in the comments.
resources: - name: sample_code type: GitRepo configuration: path: jfrog/jfrog-pipelines-matrix-example # Change to your own path branches: include: ^master$ gitProvider: jfrog_github # Change to your own GitHub integration name
PreMatrix Step
The PreMatrix generic step is an optional step for performing preamble operations. It may be used to prepare a build environment for execution of a Matrix step.
Our example PreMatrix step loads the npm package dependencies that will be needed by the steplets.
PreMatrix Example Step
- name: setup type: PreMatrix configuration: inputResources: - name: sample_code execution: onExecute: - echo "Preparing the code base" - pushd "${res_sample_code_resourcePath}/app" - npm install - popd - mkdir -p app - cp -r "${res_sample_code_resourcePath}/app/." ./app/
Matrix Step
The Matrix step executes the test cases for each module in multiple versions of the nodejs runtime. The modules are specified as environment variables in stepletMultipliers
, which also specifies the runtimes.
Matrix Example Step
- name: test type: Matrix stepMode: Bash configuration: inputSteps: - name: setup stepletMultipliers: environmentVariables: - module: mod_1 - module: mod_2 - module: mod_3 - module: mod_4 runtimes: - type: image image: auto: language: node versions: - 12 - type: image image: auto: language: node versions: - 8.17.0 execution: onExecute: - pushd ./setup/app - MODULE="$module" npm run test - popd onComplete: - save_tests ./setup/app/reports
PostMatrix Step
The PostMatrix generic step is an optional step for performing post-execution tasks following a Matrix step.
In the example pipeline, the PostMatrix step aggregates all the test reports.
PostMatrix Step Example
- name: output type: PostMatrix configuration: inputSteps: - name: test execution: onExecute: - echo "Executing the output step" - ls -ltr ./test/setup/app/reports - save_tests ./test/setup/app/reports
Results
When the example pipeline is loaded and run, the Matrix step executes eight steplets. The run log for each steplet, including saved test results, can be viewed from the pipeline's run history.
The run log of the PostMatrix step shows the aggregated results of all steplets.
Multi-node Matrix
A Matrix step can be configured such that each steplet runs on its own build node in the same node pool. This helps in reducing execution time and ensures efficient use of available resources.
This can be done by setting multiNode
to true
. Default value is false
.
Multi-node Matrix Example
steps: - name: step_1 type: Matrix stepMode: Bash configuration: multiNode: true stepletMultipliers: environmentVariables: - foo: foo - bar: bar
Multi-node pool Matrix
A Matrix step can be configured such that each steplet executes in parallel on multiple node pools with different operating systems or cloud providers.
This can be done by setting multiNode
to true
. Default value is false
.
Multi-node pool Matrix Example
steps: - name: step_1 type: Matrix stepMode: Bash configuration: multiNode: true stepletMultipliers: nodePools: - windows - ubuntu_18 - ubuntu_16 environmentVariables: - foo: foo - bar: bar