This simple CLI, written in Go, is sending jUnit metrics to a back-end using Open Telemetry.
Inspired by https://github.com/axw/test2otlp, which sends traces and spans for
go test
JSON events as they occur.
As jUnit represents a de-facto standard for test results in every programming language, this tool consumes the XML files produced by the test runner (or a tool converting to xUnit format), sending metrics to one or more open-source or commercial back-ends with Open Telemetry.
This tool will work in the context of a CI runner, such as a Github action, a Jenkins job, a Gitlab runner, or even a local execution. This is important because it will use the context of the CI execution to infer the attributes to be added to the OpenTelemetry traces and spans.
In particular the order of evaluation to detect the right execution context is the following:
Local execution > Github action > Jenkins multibranch pipeline > Gitlab runner > NIL
It reads the environment variables that are avaible in the context of a local execution, representing the fallback if no context is discovered:
// FromLocal returns an SCM context for local, using TARGET_BRANCH and BRANCH as the variables controlling
// if the SCM context represents a change request. BRANCH is mandatory, otherwise an empty context will be retrieved.
// If TARGET_BRANCH is not empty, it will represent a change request
func FromLocal() *ScmContext {
baseRef := os.Getenv("TARGET_BRANCH")
headRef := os.Getenv("BRANCH")
if headRef == "" {
return nil
}
isPR := (baseRef != "")
return &ScmContext{
ChangeRequest: isPR,
Commit: "",
Branch: headRef,
Provider: "",
TargetBranch: baseRef,
}
}
It reads the environment variables that are avaible in the context of a Github Action execution:
// FromGithub returns an SCM context for Github, reading the right environment variables, as described
// in their docs
func FromGithub() *ScmContext {
if os.Getenv("GITHUB_SHA") == "" {
return nil
}
sha := os.Getenv("GITHUB_SHA")
branchName := os.Getenv("GITHUB_REF_NAME")
baseRef := os.Getenv("GITHUB_BASE_REF") // only present for pull requests on Github Actions
headRef := os.Getenv("GITHUB_HEAD_REF") // only present for pull requests on Github Actions
isChangeRequest := (baseRef != "" && headRef != "")
return &ScmContext{
ChangeRequest: isChangeRequest,
Commit: sha,
Branch: branchName,
Provider: "Github",
TargetBranch: baseRef,
}
}
It reads the environment variables that are avaible in the context of a Jenkins multibranch pipeline execution:
// FromJenkins returns an SCM context for Jenkins, reading the right environment variables, as described
// in their docs
func FromJenkins() *ScmContext {
if os.Getenv("JENKINS_URL") == "" {
return nil
}
isPR := os.Getenv("CHANGE_ID") != "" // only present on multibranch pipelines on Jenkins
headRef := os.Getenv("BRANCH_NAME") // only present on multibranch pipelines on Jenkins
sha := os.Getenv("GIT_COMMIT") // only present on multibranch pipelines on Jenkins
baseRef := os.Getenv("CHANGE_TARGET") // only present on multibranch pipelines on Jenkins
if isPR {
return &ScmContext{
ChangeRequest: isPR,
Commit: sha,
Branch: headRef,
Provider: "Jenkins",
TargetBranch: baseRef,
}
} else {
return &ScmContext{
ChangeRequest: isPR,
Commit: sha,
Branch: headRef,
Provider: "Jenkins",
TargetBranch: headRef,
}
}
}
It reads the environment variables that are avaible in the context of a Gitlab runner execution:
// FromGitlab returns an SCM context for Gitlab, reading the right environment variables, as described
// in their docs
func FromGitlab() *ScmContext {
if os.Getenv("CI_COMMIT_REF_NAME") == "" {
return nil
}
sha := os.Getenv("CI_MERGE_REQUEST_SOURCE_BRANCH_SHA") // only present on merge requests on Gitlab CI
commitBranch := os.Getenv("CI_COMMIT_BRANCH") // only present on branches on Gitlab CI
headRef := os.Getenv("CI_COMMIT_REF_NAME") // only present on branches on Gitlab CI
baseRef := os.Getenv("CI_MERGE_REQUEST_TARGET_BRANCH_NAME") // only present on merge requests on Gitlab CI
isChangeRequest := (commitBranch == "")
return &ScmContext{
ChangeRequest: isChangeRequest,
Commit: sha,
Branch: headRef,
Provider: "Gitlab",
TargetBranch: baseRef,
}
}
This tool is able to override the following attributes:
Attribute | Flag | Default value | Description |
---|---|---|---|
Repository Path | --repository-path | . |
Path to the SCM repository to be read. |
Service Name | --service-name | junit2otlp |
Overrides OpenTelemetry's service name. If the OTEL_SERVICE_NAME environment variable is set, it will take precedence over any other value. |
Service Version | --service-version | Empty | Overrides OpenTelemetry's service version. If the OTEL_SERVICE_VERSION environment variable is set, it will take precedence over any other value. |
Trace Name | --trace-name | junit2otlp |
Overrides OpenTelemetry's trace name. |
For using this tool in a distributed tracing scenario, where there is a parent trace in which the test reports traces should be attached, it's important to set the TRACEPARENT
environment variable, so that the traces and spans generated by this tool are located under the right parent trace. Please read more on this here.
For further reference on environment variables in the OpenTelemetry SDK, please read the official specification
This tool is going to parse the XML report produced by jUnit, or any other tool converting to that format, adding different attributes, separated by different categories:
- Test metrics attributes
- Ownership attributes
The following attributes are added as metrics and/or traces.
For each test execution, represented by a test report file, the tool will add the following attributes to the metric document, including them in the trace representing the test execution.
Attribute | Description |
---|---|
tests.suite.failed |
Number of failed tests in the test execution |
tests.suite.error |
Number of errored tests in the test execution |
tests.suite.passed |
Number of passed tests in the test execution |
tests.suite.skipped |
Number of skipped tests in the test execution |
tests.suite.duration |
Duration of the test execution |
tests.suite.suitename |
Name of the test execution |
tests.suite.systemerr |
Log produced by Systemerr |
tests.suite.systemout |
Log produced by Systemout |
tests.suite.total |
Total number of tests in the test execution |
For each test case in the test execution, the tool will add the following attributes to the span document representing the test case:
Attribute | Description |
---|---|
tests.case.classname |
Classname or file for the test case |
tests.case.duration |
Duration of the test case |
tests.case.error |
Error message of the test case |
tests.case.message |
Message of the test case |
tests.case.status |
Status of the test case |
tests.case.systemerr |
Log produced by Systemerr |
tests.case.systemout |
Log produced by Systemout |
These attributes are added to the traces and spans sent by the tool, identifying the owner (or owners) of the test suite, trying to correlate a test failure with an author or authors. To identify the owner, the tool will inspect the SCM repository for the project.
Because the XML test report is evaluated for a project in a SCM repository, the tool will add the following attributes to each trace and span:
Attribute | Description |
---|---|
scm.authors |
Array of unique Email addresses for the authors of the commits |
scm.baseRef |
Name of the target branch (Only for change requests) |
scm.branch |
Name of the branch where the test execution is processed |
scm.committers |
Array of unique Email addresses for the committers of the commits |
scm.provider |
Optional. If present, will include the name of the SCM provider, such as Github, Gitlab, Bitbucket, etc. |
scm.repository |
Array of unique URLs representing the repository (i.e. https://github.com/mdelapenya/junit2otlp) |
scm.type |
Type of the SCM (i.e. git, svn, mercurial) At this moment the tool only supports Git repositories. |
The tool will add the following attributes to each trace and span if and only if the XML test report is evaluated in the context of a change requests for a Git repository:
Attribute | Description |
---|---|
scm.git.additions |
Number of added lines in the changeset |
scm.git.deletions |
Number of deleted lines in the changeset |
scm.git.clone.depth |
Depth of the git clone |
scm.git.clone.shallow |
Whethere the git clone was shallow or not |
scm.git.files.modified |
Number of modified files in the changeset |
A changeset is calculated based on the HEAD commit and the first ancestor between HEAD and the branch where the changeset is submitted against.
It's possible to run the binary as a Docker image. To build and use the image
- First build the Docker image using this Make goal:
make build-docker-image
- Then start the Elastic Stack back-end:
make demo-start-elastic
- Finally, once the services are started, run:
cat TEST-sample3.xml | docker run --rm -i --network elastic_junit2otlp --volume "$(pwd):/opt/projectname" --env OTEL_EXPORTER_OTLP_ENDPOINT=http://apm-server:8200 mdelapenya/junit2otlp:latest --service-name DOCKERFOO --trace-name TRACEBAR --repository-path "/opt/projectname"
- We are making the Docker container receive the pipe with the
-i
flag. - We are attaching the container to the same Docker network where the services are running.
- We are passing an environment variable with the URL of the OpenTelemetry exporter endpoint, in this case an APM Server instance.
- We are passing command line flags to the container, setting the service name (DOCKERFOO) and the trace name (TRACEBAR).
To demonstrate how traces and metrics are sent to different back-ends, we are provising the following demos:
- Elastic
- Jaeger
- Prometheus
- Zipkin
It will use the Elastic Stack as back-end, sending the traces, spans and metrics through the APM Server, storing them in Elasticsearch and finally using Kibana as visualisation layer.
make demo-start-elastic
go build && chmod +x ./junit2otlp
cat TEST-sample.xml | ./junit2otlp
cat TEST-sample2.xml | ./junit2otlp
cat TEST-sample3.xml | ./junit2otlp
open http://localhost:5601/app/apm/services?rangeFrom=now-15m&rangeTo=now&comparisonEnabled=true&comparisonType=day
It will use Jaeger as back-end, sending the traces, spans and metrics through the OpenTelemetry collector, storing them in memory.
make demo-start-jaeger
go build && chmod +x ./junit2otlp
cat TEST-sample.xml | ./junit2otlp
cat TEST-sample2.xml | ./junit2otlp
cat TEST-sample3.xml | ./junit2otlp
open http://localhost:16686
It will use Prometheus as back-end, sending the traces, spans and metrics through the OpenTelemetry collector, storing them in memory.
make demo-start-prometheus
go build && chmod +x ./junit2otlp
cat TEST-sample.xml | ./junit2otlp
cat TEST-sample2.xml | ./junit2otlp
cat TEST-sample3.xml | ./junit2otlp
open http://localhost:9090
It will use Prometheus as back-end, sending the traces, spans and metrics through the OpenTelemetry collector, storing them in memory.
make demo-start-zipkin
go build && chmod +x ./junit2otlp
cat TEST-sample.xml | ./junit2otlp
cat TEST-sample2.xml | ./junit2otlp
cat TEST-sample3.xml | ./junit2otlp
open http://localhost:9411