Test Visibility Troubleshooting

Cette page n'est pas encore disponible en français, sa traduction est en cours.
Si vous avez des questions ou des retours sur notre projet de traduction actuel, n'hésitez pas à nous contacter.

CI Visibility is not available in the selected site () at this time.


This page provides information to help you troubleshot issues with Test Visibility. If you need additional help, contact Datadog Support.

Your tests are instrumented, but Datadog isn’t showing any data

  1. Go to the Tests page for the language you’re instrumenting and check that the testing framework you are using is supported in the Compatibility section.
  2. Check if you see any test results in the Test Runs section. If you do see results there, but not in the Tests section, Git information is missing. See Data appears in Test Runs but not Tests to troubleshoot it.
  3. If you are reporting the data through the Datadog Agent, make sure it is running on the host where tests are run (accessible at localhost:8126), or if accessible on another hostname or port, make sure you run your tests with the appropriate Agent hostname set in the DD_AGENT_HOST and the appropriate port in DD_TRACE_AGENT_PORT environment variables. You can activate debug mode in the tracer to check if it’s able to connect to the Agent.
  4. If you still don’t see any results, contact Support for troubleshooting help.

You are uploading JUnit test reports with datadog-ci but some or all tests are missing

If you are uploading JUnit test report files with datadog-ci CLI and you do not see the tests, it is likely the tests are being discarded because the report is considered incorrect.

The following aspects make a JUnit test report incorrect:

  • A timestamp of the reported tests that is older than 71 hours before the moment the report is uploaded.
  • A testsuite without a name.

Data appears in test runs but not tests

If you can see test results data in the Test Runs tab, but not the Tests tab, Git metadata (repository, commit, or branch) is probably missing. To confirm this is the case, open a test execution in the Test Runs section, and check that there is no git.repository_url, git.commit.sha, or git.branch. If these tags are not populated, nothing shows in the Tests section.

  1. Tracers first use the environment variables, if any, set by the CI provider to collect Git information. See Running tests inside a container for a list of environment variables that the tracer attempts to read for each supported CI provider. At a minimum, this populates the repository, commit hash, and branch information.

  2. Next, tracers fetch Git metadata using the local .git folder, if present, by executing git commands. This populates all Git metadata fields, including commit message, author, and committer information. Ensure the .git folder is present and the git binary is installed and in $PATH. This information is used to populate attributes not detected in the previous step.

  3. You can also provide Git information manually using environment variables, which override information detected by any of the previous steps.

    The supported environment variables for providing Git information are:

    DD_GIT_REPOSITORY_URL (required)
    URL of the repository where the code is stored. Both HTTP and SSH URLs are supported.
    Example: git@github.com:MyCompany/MyApp.git, https://github.com/MyCompany/MyApp.git
    DD_GIT_COMMIT_SHA (required)
    Full (40-character long SHA1) commit hash.
    Example: a18ebf361cc831f5535e58ec4fae04ffd98d8152
    Git branch being tested. Leave empty if providing tag information instead.
    Example: develop
    Git tag being tested (if applicable). Leave empty if providing branch information instead.
    Example: 1.0.1
    Commit message.
    Example: Set release number
    Commit author name.
    Example: John Smith
    Commit author email.
    Example: john@example.com
    Commit author date in ISO 8601 format.
    Example: 2021-03-12T16:00:28Z
    Commit committer name.
    Example: Jane Smith
    Commit committer email.
    Example: jane@example.com
    Commit committer date in ISO 8601 format.
    Example: 2021-03-12T16:00:28Z
  4. If no CI provider environment variables are found, tests results are sent with no Git metadata.

The tests wall time is empty

If you cannot see the tests wall time it is likely that the CI provider metadata is missing. To confirm this is the case, open a test execution in the Test Runs section, and check if the ci.pipeline.id, ci.pipeline.name, ci.pipeline.number, or ci.job.url tags are missing. If these tags are not populated, then nothing shows in the wall time column.

  1. Tracers use the environment variables set by the CI provider to collect this information. See Running tests inside a container for a list of environment variables that the tracer attempts to read for each supported CI provider. Make sure that the environment variables have the expected values set.
  2. Check that you are running your tests in a supported CI provider. For a list of supported CI providers, see Running tests inside a container. Only these CI providers can extract the information to enrich the test metadata with CI information.
  3. If you still don’t see the wall time, contact Datadog support for help.

The tests wall time is not what is expected

How wall time is calculated

The wall time is defined as the time difference between the start time of the first test and the end time of the last test for the given pipeline.

This is done using the following algorithm:

  1. Compute a hash based on CI information to group the tests.
    1. If the tests include ci.job.url, use this tag to calculate the hash.
    2. If the tests don’t include ci.job.url, use ci.pipeline.id + ci.pipeline.name + ci.pipeline.number to calculate the hash.
  2. The calculated wall time is associated to a given hash. Note: If there are multiple jobs that execute tests, the wall time is calculated for each job, and the maximum from all calculated wall times is shown.

Possible issues with wall time calculation

If you’re using a library for testing time-dependent code, like timecop for Ruby or FreezeGun for Python, it is possible that test timestamps are wrong, and therefore calculated wall times. If this is the case, make sure that modifications to time are rolled back before finishing your tests.

The test status numbers are not what is expected

The test status numbers are calculated based on the unique tests that were collected. The uniqueness of a test is defined not only by its suite and name, but by its test parameters and test configurations as well.

The numbers are lower than expected

If the numbers are lower than expected, it is likely that either the library or the tool you are using to collect test data cannot collect test parameters and/or some test configurations.

  1. If you are uploading JUnit test report files:
    1. If you are running the same tests in different environment configurations, make sure you are setting those configuration tags during the upload.
    2. If you are running parameterized tests, it’s very likely that the JUnit report does not have that information. Try using a native library to report test data.
  2. If you still don’t see the expected results, contact Datadog support for troubleshooting help.

The passed/failed/skipped numbers are different than expected

If the same test is collected several times for the same commit but with different status, the aggregated result follows the algorithm in the table below:

Test Status - First TryTest Status - Retry #1Result

The default branch is not correct

How it impacts the product

The default branch is used to power some features of the products, namely:

  • Default branches list on the Tests page: This list only displays default branches. Setting the wrong default branch can result in missing or incorrect data in the default branches list.

  • Wall time comparison for non-default branches: On the Tests page, in the Branches view, the VS Default column is calculated by comparing wall time for the current branch against wall time for the default branch.

  • New flaky tests: Tests that are not currently classified as flaky in the default branch. If the default branch is not properly set, this could lead to a wrong number of detected new flaky tests.

  • Pipelines list: The pipelines list only displays default branches. Setting the wrong default branch can result in missing or incorrect data in the pipelines list.

How to fix the default branch

If you have admin access, you can update it from the Repository Settings Page.

Execution history is not available for a specific test case

Other symptoms of the same issue include:

  • A test case is not classified as flaky even if it exhibits flakiness.
  • A test case cannot be skipped by Intelligent Test Runner.

It is likely that the test case configuration is unstable because one or more of the test parameters are non-deterministic (for instance, they include current date or a random number).

The best way to fix this is to make sure that the test parameters are the same between test runs.

Further reading

Documentation, liens et articles supplémentaires utiles: