CI Build ID

Using --ci-build-id flag for Playwright and Cypress tests parallelization

What is CI Build ID?

CI Build ID is a unique identifier used by Currents to distinguish between different runs. Multiple CI machines using the same CI Build ID will report the results to the same run.

CI Build ID and Parallelization

Using unique CI Build ID in different builds

Imagine a CI pipeline running tests in parallel using multiple machines. Starting two builds with a different CI Build ID will create 2 distinct "Runs" in Currents dashboard.

The parallelization and reporting will happen for each build independently from the other. That is usually the desired situation - each build should have a unique CI Build ID.

Using the same CI Build ID in different builds

In contrast, consider a situation when 2 different builds use the same CI Build ID. That's an uncommon situation, but it's worth demonstrating for understanding the use of CI Build ID.

We created two different builds with the same CI Build ID. That will result in 6 machines executing the same parallelized set of tests and reporting their results to the same run.

If Cypress Parallelization or Playwright Orchestration are enabled, all the spec files will be distributed between those machines.

Generating a unique CI Build ID

Currents Dashboard generates CI Build ID automatically for popular CI environments, based on environment variables.

However, you can specify CI Build ID explicitly using --ci-build-id flag, for example:

currents run --ci-build-id $BRANCH_$BUILDID --parallel --record --key xxx 

In order to manually construct a CI Build ID that is unique for each build (but similar across all the participating machines) it is recommended to use your CI provider's environment variables that combine pipeline/job/build identifier and also an attempt number.

For example, for GitHub Actions:

"${{ github.repository }}-${{ github.run_id }}-${{ github.run_attempt }}"

Here's a list of popular providers and the environment variables that can be used to construct CI Build ID.

CI ProviderEnv Variable

AWS CodeBuild

CODEBUILD_INITIATOR

Bitbucket

BITBUCKET_BUILD_NUMBER

GitHub Actions

RUN_ID +

RUN_ATTEMPT

CircleCI

CIRCLE_WORKFLOW_ID + CIRCLE_BUILD_NUM

GitLab

CI_PIPELINE_ID

Jenkins

BUILD_NUMBER

Travis

TRAVIS_BUILD_ID

Refer to your CI provider documentation for the list of available environment variables. Also, see the official cypress documentation on CI Build ID.

FAQ: Why does each CI machine run all the tests?

One popular and confusing scenario is:

  • the first build completes all the tests

  • the second build uses the same CI Build ID and immediately finishes without running any test at all

That's because both builds use the same CI Build ID - the second build "joins" an already finished run that has no more tests to execute.

In most chances, each CI machine generates a different CI Build ID. Each unique CI Build ID creates a new run and executes all the tests. Please make sure that you provide the same CI Build ID across different CI machines that are part of the same build.

FAQ: Retrying a build doesn't run cypress tests at all

Most chances you're reusing a CI Build ID for a run that was already completed. In order to create a new run, please use a new, unique CI Build ID.

FAQ: Retrying builds and CI Build ID

Imagine a situation

  • You start a new build with CI Build ID build-001

    • Build completes and reports all the results to Currents Dashboard

    • Currents marks build-001 as "finished" and all the files as completed

  • You restart the build (attempt B), but keep the same CI build ID build-001

    • Currents considers build-001 as already completed

    • Currents won't accept new results for build-001, because all the results were already reported

    • Currents will not send any new files for Cypress orchestration, because build-001 already ran all the spec files on the first attempt

To resolve this ambiguity, we need to have a different CI build ID for each rerun.

Most CI providers provide a different set of environment variables for different attempts and Currents dashboard can identify it automatically - it will create an entirely new run for retries.

You can also construct an explicit CI Build ID when retrying a build, for example, for GitHub Actions:

"${{ github.repository }}-${{ github.run_id }}-${{ github.run_attempt }}"

If you are generating CI Build ID manually, please make sure to include the retry/attempt identifier.

Please refer to your CI tool documentation to explore what environment variables are available for composing a valid CI Build ID.

FAQ: Using commit SHA as CI Build ID

Using commit SHA as a CI Build ID is a valid approach and can work for many setups. However, please be aware that rerunning a build with the same commit SHA can result in a duplicate CI Build ID and prevent orchestration and reporting (see FAQ: Retrying builds and CI Build ID)

FAQ: Retrying a build only for failed tests

TL;DR Currents Dashboard will always run all the tests using the available machines, even for reruns. That's due to the architectural limitations of load balancing.

Some CI providers (e.g. GitLab, GitHub) allow reruns only for the failed containers. Invoking such a rerun will result in:

  • a unique CI Build ID would be generated

  • it will create a completely new run within the dashboard

  • the dashboard will load balance all the specs among all the available containers

So, you end up running all the tests using a just subset of available containers.

We have been experimenting with alternative load-balancing strategies that would allow seamless reruns. Please reach out to our customer support if you want to get updates regarding the progress.

Please note: GitLab does not provide a "rerun identifier" within its CI environment. See the WIP discussion.

Last updated