prow-ci

name	prow-ci
description	Fetch and analyze OpenShift Prow CI job failures with automated artifact download and failure pattern detection
trigger	prow, prow-ci, /prow-ci, ci results, check ci, analyze ci failure

Prow CI Analysis for Managed Velero Operator

This skill fetches Prow CI job artifacts from Google Cloud Storage and provides automated failure analysis.

Prerequisites

Before using this skill, verify gcloud CLI is installed:

which gcloud

If not installed, provide instructions from: https://cloud.google.com/sdk/docs/install

Note: The test-platform-results GCS bucket is publicly accessible - no authentication required.

Quick Start

# Check PR status and get Prow job URLs
gh pr checks <PR_NUMBER>

# Analyze a failed job
/prow-ci <prow-job-url>

# Or ask naturally:
"Analyze the lint failure in PR <NUMBER>"
"Check why the validate job failed"
"Show me what broke in the coverage job"

Implementation

When invoked, this skill:

1. Fetches artifacts using fetch_prow_artifacts.py:

   • Downloads prowjob.json (job metadata)
   • Downloads build-log.txt (complete build output with all errors)
   • Saves to .work/prow-artifacts/<build-id>/
   • Note: Script is optimized to only download essential files. Optional artifacts (JUnit XML, per-target logs) are skipped as build-log.txt contains all needed information.

2. Analyzes failures using analyze_failure.py:

   • Parses build-log.txt for error patterns
   • Detects common failure patterns (lint, build, timeout, OOM)
   • Extracts error messages and stack traces
   • Identifies compilation errors and test failures

3. Generates report:

   • Markdown format with failure summary
   • Pattern detection (compilation errors, lint failures, timeouts)
   • Top error messages and failures
   • Actionable failure details

Usage Instructions

Step 1: Get Prow Job URL

# View PR checks to find failed jobs
gh pr checks <PR_NUMBER>

# Or get detailed status
gh pr view <PR_NUMBER> --json statusCheckRollup --jq '.statusCheckRollup[] | select(.state == "FAILURE")'

Example Prow job URL:

https://prow.ci.openshift.org/view/gs/test-platform-results/pr-logs/pull/openshift_managed_velero_operator/<PR_NUMBER>/pull-ci-openshift-managed-velero-operator-master-lint/<BUILD_ID>

Step 2: Fetch and Analyze

Run the fetch script from repository root:

# From repository root
python3 .claude/skills/prow-ci/fetch_prow_artifacts.py "<prow-job-url>" -o .work/prow-artifacts

This downloads only the essential files:

• prowjob.json - Job metadata (job name, state, type, URL)
• build-log.txt - Complete build output (contains all errors, test failures, and output)

Step 3: Analyze Failures

python3 .claude/skills/prow-ci/analyze_failure.py .work/prow-artifacts/<build-id> -f markdown

Output includes:

• Job information (name, state, URL)
• Detected failure patterns (lint errors, build failures, timeouts)
• Top error messages from build log
• Failure details extracted from log

Step 4: Present Findings

Create a clear summary for the user with:

• Root cause identification
• Detected patterns (lint, build, timeout, etc.)
• Key error messages
• Actionable next steps to fix the issue

Example Workflow

# User provides: "Analyze the lint failure in PR <NUMBER>"

# 1. Get Prow job URL
gh pr checks <PR_NUMBER> | grep lint

# 2. Fetch artifacts
python3 .claude/skills/prow-ci/fetch_prow_artifacts.py \
  "https://prow.ci.openshift.org/view/gs/test-platform-results/pr-logs/pull/openshift_managed_velero_operator/<PR_NUMBER>/pull-ci-openshift-managed-velero-operator-master-lint/<BUILD_ID>"

# 3. Analyze
python3 .claude/skills/prow-ci/analyze_failure.py \
  .work/prow-artifacts/<BUILD_ID> \
  -f markdown

# 4. Review the output and provide actionable summary

Prow Resources

Main Dashboard: https://prow.ci.openshift.org/
CI Search: https://github.com/openshift/ci-search
Job History: https://prow.ci.openshift.org/?repo=openshift%2Fmanaged-velero-operator

Common Use Cases

1. Check Recent CI Results

# Check latest job status for specific PR
gh pr view PR_NUMBER --json statusCheckRollup --jq '.statusCheckRollup[] | select(.context | contains("prow"))'

# Or view all checks for a PR
gh pr checks PR_NUMBER

2. Access Build Logs

Prow logs are stored at:

• Pull request jobs: gs://test-platform-results/pr-logs/pull/openshift_managed_velero_operator/[PR_NUMBER]/[JOB_NAME]/[JOB_ID]
• Periodic jobs: gs://test-platform-results/logs/[JOB_NAME]/[JOB_ID]

Viewing logs via web:

https://prow.ci.openshift.org/view/gs/test-platform-results/pr-logs/pull/openshift_managed_velero_operator/[PR_NUMBER]/[JOB_NAME]/[JOB_ID]

3. Analyze Test Failures

# Get PR checks
gh pr view PR_NUMBER --json statusCheckRollup

# Find failed jobs
gh pr checks PR_NUMBER | grep -i "fail"

# Use this skill to fetch and analyze failures
# The script downloads prowjob.json and build-log.txt for analysis

4. Common Job Names

Prow CI Jobs (configured in openshift/release):

• pull-ci-openshift-managed-velero-operator-master-e2e-binary-build-success - E2E binary build verification
• pull-ci-openshift-managed-velero-operator-master-coverage - Code coverage analysis (with Codecov)
• pull-ci-openshift-managed-velero-operator-master-lint - Linting checks
• pull-ci-openshift-managed-velero-operator-master-test - Unit tests
• pull-ci-openshift-managed-velero-operator-master-validate - Validation checks

Tekton Pipelines (configured in .tekton/):

• managed-velero-operator-pull-request - Main PR pipeline (docker build with OCI-TA)
• managed-velero-operator-e2e-pull-request - E2E testing pipeline
• managed-velero-operator-pko-pull-request - PKO (Package Operator) pipeline
• Corresponding -push pipelines for merged commits

Debugging CI Failures

Step 1: Identify Failed Job

gh pr checks PR_NUMBER

Step 2: Access Prow UI

Open the Prow link from PR checks or construct manually:

https://prow.ci.openshift.org/?repo=openshift%2Fmanaged-velero-operator&type=presubmit

Step 3: Review Logs

Click on failed job → "Build Log" tab

Step 4: Check Artifacts

Look for:

• Test failure logs
• Coverage reports
• Generated artifacts

Step 5: Reproduce Locally

Many Prow jobs can be reproduced with:

# For unit tests (matches: pull-ci-...-test)
make go-test

# For linting (matches: pull-ci-...-lint)
make go-check
# OR use prek for comprehensive linting
prek run --all-files

# For validation (matches: pull-ci-...-validate)
make validate

# For coverage (matches: pull-ci-...-coverage)
make coverage

# For E2E binary build (matches: pull-ci-...-e2e-binary-build-success)
make e2e-binary-build

# For container builds (Tekton pipelines)
make docker-build

CI/Prow Integration in This Repo

This repo uses both Prow and Tekton for comprehensive CI:

Prow CI (openshift/release):

• Configuration: ci-operator/config/openshift/managed-velero-operator/openshift-managed-velero-operator-master.yaml
• Runs: lint, test, validate, coverage, e2e-binary-build
• Uses Codecov for coverage reporting (secret: managed-velero-operator-codecov-token)
• Skip rules: Changes to .tekton/, GitHub (.github/), .md files, OWNERS, LICENSE don't trigger most jobs

Tekton Pipelines (.tekton/):

• Primary build pipeline using Pipelines as Code
• Three pipeline types: main, e2e, pko
• Builds container images to Quay (managed-velero-operator-tenant)
• Pull request images expire after 5 days
• Uses boilerplate framework from openshift/boilerplate (docker-build-oci-ta pipeline)

Quick Reference Commands

# Check all PR checks status
gh pr checks <PR_NUMBER>

# View detailed status for a specific PR
gh pr view <PR_NUMBER> --json statusCheckRollup

# Filter only Prow jobs
gh pr checks <PR_NUMBER> | grep "pull-ci-openshift-managed-velero-operator"

# Check Tekton pipeline status
gh pr view <PR_NUMBER> --json statusCheckRollup --jq '.statusCheckRollup[] | select(.context | contains("Tekton"))'

# Open Prow dashboard in browser (cross-platform)
# Copy and paste this URL into your browser:
# https://prow.ci.openshift.org/?repo=openshift%2Fmanaged-velero-operator

# Or use platform-specific command:
# macOS: open "https://prow.ci.openshift.org/?repo=openshift%2Fmanaged-velero-operator"
# Linux: xdg-open "https://prow.ci.openshift.org/?repo=openshift%2Fmanaged-velero-operator"
# Windows: start "https://prow.ci.openshift.org/?repo=openshift%2Fmanaged-velero-operator"

# View specific PR on Prow (replace <PR_NUMBER>)
# https://prow.ci.openshift.org/?repo=openshift%2Fmanaged-velero-operator&type=presubmit&pull=<PR_NUMBER>

Troubleshooting

Can't find job results?

• Check both Prow AND Tekton - this repo uses both systems
• Prow jobs: pull-ci-openshift-managed-velero-operator-master-*
• Tekton jobs: Usually show as "Tekton" or pipeline names in PR checks
• Verify repo name format in Prow: openshift_managed_velero_operator (underscore, not dash)
• Ensure PR has been opened and CI has run

Logs show permission denied?

• Prow logs are public for openshift org
• Use web UI (prow.ci.openshift.org) instead of gsutil
• Check if job ID is correct

Job still running?

• Check Prow dashboard for in-progress jobs
• Look for "Pending" or "Running" status
• Wait for completion before accessing artifacts

Tekton pipeline failures?

• Check the pipeline link in PR checks (usually links to Konflux/AppStudio UI)
• Tekton logs are in the AppStudio dashboard, not Prow
• Common issues:
  • Image build failures → Check Dockerfile syntax and build context
  • Pipeline timeout → Check for slow steps or network issues
  • Auth failures → Secret configuration in managed-velero-operator-tenant namespace
• Local validation:

  # Validate Tekton YAML syntax
  kubectl apply --dry-run=client -f .tekton/
  
  # Test container build locally
  podman build -f build/Dockerfile -t test:local .
  

Advanced: CI Search

For historical job searches:

# Clone ci-search tool
git clone https://github.com/openshift/ci-search.git

# Use web interface at search.ci.openshift.org (if available)
# Search for patterns in build logs across all jobs

References

• Prow Dashboard
• CI Search Tool
• OpenShift CI Documentation

CI Configuration Files

Prow Configuration (in openshift/release repo):

• Location: ci-operator/config/openshift/managed-velero-operator/openshift-managed-velero-operator-master.yaml
• Update process: Submit PR to openshift/release repository
• Auto-generated jobs in: ci-operator/jobs/openshift/managed-velero-operator/

Tekton Pipelines (in this repo):

• Location: .tekton/ directory
• Files:
  • managed-velero-operator-pull-request.yaml - Main PR pipeline
  • managed-velero-operator-push.yaml - Post-merge pipeline
  • managed-velero-operator-e2e-pull-request.yaml - E2E testing
  • managed-velero-operator-pko-pull-request.yaml - PKO validation
• Triggered by: Pipelines as Code (via Tekton)
• Uses: Boilerplate docker-build-oci-ta pipeline from openshift/boilerplate

Coverage Reporting

This repository uses Codecov for coverage tracking:

• Secret: managed-velero-operator-codecov-token (stored in Prow)
• Generate coverage locally: make coverage
• Coverage runs on PRs and post-merge (publish-coverage)
• Dashboard: Check Codecov for managed-velero-operator

Integration with Other Skills

• Use with test-agent to compare local test results with CI
• Use with ci-agent to validate CI configuration
• Use with lint-agent when investigating lint failures in CI
• Use with security-agent when investigating pre-commit hook failures