upload job info if debug mode set #374
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| <<<<<<< Updated upstream | ||
| # This workflow is meant to imitate the behavior of RAPIDS project PR workflows, such as | ||
| on: | ||
| pull_request: | ||
| workflow_dispatch: | ||
| env: | ||
| SHARED_ACTIONS_REF: ${{ github.ref}} | ||
| defaults: | ||
| run: | ||
| shell: bash | ||
| jobs: | ||
| base-env-setup: | ||
| runs-on: ubuntu-latest | ||
| # These will be stashed. The names are not arbitrary. They match special OpenTelemetry names | ||
| # or names that are hard-coded in actions/scripts downstream. | ||
| env: | ||
| SHARED_ACTIONS_REPO: rapidsai/shared-actions | ||
| SHARED_ACTIONS_REF: ${{ github.ref }} | ||
| # this should stay the same throughout this workflow, but child workflows will each | ||
| # have their own OTEL_SERVICE_NAME. It is generally the job name, including any matrix elements. | ||
| # This is what distinguishes one job trace from another, so it is important to be distinct | ||
| # between jobs. | ||
| OTEL_SERVICE_NAME: test-telemetry plus something | ||
| # TODO: this should be set as an org-wide variable | ||
| OTEL_EXPORTER_OTLP_ENDPOINT: https://tempo.gha-runners.nvidia.com:4318 | ||
| OTEL_EXPORTER_OTLP_PROTOCOL: "http/protobuf" | ||
| OTEL_RESOURCE_ATTRIBUTES: "git.repository=${{ github.repository }},git.ref=${{ github.ref }},git.sha=${{ github.sha }},git.job_url=${{ github.server_url }}/${{ github.repository }}/actions/runs/${{ github.run_id }}" | ||
| outputs: | ||
| service-name: ${{ steps.export.outputs.service_name }} | ||
| steps: | ||
| - name: Compute traceparent and stash telemetry-related env vars | ||
| uses: rapidsai/shared-actions/telemetry-dispatch-stash-base-env-vars@telemetry-dispatch-actions | ||
| - name: Export service name so we can check it below | ||
| id: export | ||
| run: echo service_name="${OTEL_SERVICE_NAME}" >> ${GITHUB_OUTPUT} | ||
| child-workflow: | ||
| needs: base-env-setup | ||
| secrets: inherit | ||
| uses: rapidsai/shared-actions/.github/workflows/test-child-workflow.yaml@telemetry-dispatch-actions | ||
| summarize-top-level: | ||
| runs-on: ubuntu-latest | ||
| continue-on-error: true | ||
| needs: | ||
| - base-env-setup | ||
| - child-workflow | ||
| steps: | ||
| - name: Load base env vars, including OTEL_SERVICE_NAME | ||
| uses: rapidsai/shared-actions/telemetry-dispatch-load-base-env-vars@telemetry-dispatch-actions | ||
| with: | ||
| load_service_name: "true" | ||
| - name: Check if service name took on an unexpected value | ||
| run: | | ||
| echo "(should be the value set to the OTEL_SERVICE_NAME env var in base-env-setup job)" | ||
| [ "${OTEL_SERVICE_NAME}" = "${{needs.base-env-setup.outputs.service-name}}" ] || exit 1 | ||
| - name: Telemetry summarize | ||
| uses: rapidsai/shared-actions/telemetry-dispatch-write-summary@telemetry-dispatch-actions | ||
| with: | ||
| cert_concat: "${{ secrets.OTEL_EXPORTER_OTLP_CA_CERTIFICATE }};${{ secrets.OTEL_EXPORTER_OTLP_CLIENT_CERTIFICATE }};${{ secrets.OTEL_EXPORTER_OTLP_CLIENT_KEY }}" | ||
| - name: Check if service name was altered during telemetry summary | ||
| run: | | ||
| echo "(should be the value set to the OTEL_SERVICE_NAME env var in base-env-setup job)" | ||
| [ "${OTEL_SERVICE_NAME}" = "${{needs.base-env-setup.outputs.service-name}}" ] || exit 1 | ||
| - name: Query the Tempo HTTP API and check that our trace is present and has expected properties | ||
| run: | | ||
| TRACE_ID=$( cut -d '-' -f 2 <<< "$TRACEPARENT" ); | ||
| echo "Trace ID is: ${TRACE_ID}"; | ||
| TRACE_URL="${OTEL_EXPORTER_OTLP_ENDPOINT/4318/3200}/api/traces/${TRACE_ID}" | ||
| echo "Trace URL is: ${TRACE_URL}" | ||
| curl \ | ||
| --cert /tmp/certs/client.crt.pem --key /tmp/certs/client.key.pem --cacert /tmp/certs/ca.crt.pem \ | ||
| -Gs "${TRACE_URL}" > trace_record.json; | ||
| - name: Upload trace record | ||
| uses: actions/upload-artifact@v4 | ||
| with: | ||
| name: trace-record | ||
| path: trace_record.json | ||
| - name: Validate span metadata | ||
| # these are not returned in any particular order. The span kind is the only one | ||
| # that we can reliably expect to be the same. | ||
| run: | | ||
| span_kind="$(jq -r '.batches[0].scopeSpans[0].spans[0].kind' trace_record.json )"; | ||
| echo "Checking if span kind is as expected" | ||
| echo "Span kind is: \"${span_kind}\"" | ||
| [ "${span_kind}" = "SPAN_KIND_CLIENT" ] || exit 1 | ||
| echo "Verify that job names (also called service name) are correct" | ||
| job_names="$(jq -c '[.batches[].resource.attributes[] | select(.key == "service.name") | .value.stringValue] | unique' trace_record.json)" | ||
| [ "$job_names" = '["child-workflow / Jobby McJobface (with)","child-workflow / Jobby McJobface (without)","test-telemetry plus something"]' ] || exit 1 | ||
| ======= | ||
| >>>>>>> Stashed changes | ||
