Debugging Pipeline Failures Skill

A systematic approach to investigating and resolving Konflux Tekton pipeline failures using standard Kubernetes and Tekton tools.

What This Skill Does

This skill teaches Claude Code to:

• Follow systematic methodology - 5-phase investigation process from identification to root cause
• Use standard tools - kubectl and Tekton CLI commands for debugging
• Correlate multiple sources - Logs, events, and resource states for complete picture
• Distinguish cause from symptom - Identify root causes, not just surface-level issues
• Provide actionable guidance - Specific kubectl commands and troubleshooting steps

When to Use

Invoke this skill when dealing with:

• Failed PipelineRuns or TaskRuns in Konflux
• Build failures or test errors
• Stuck or hanging pipelines
• Image pull issues (ImagePullBackOff)
• Resource exhaustion (OOMKilled, quota errors)
• Timeout failures
• Workspace or volume mount problems
• ServiceAccount permission errors

Key Features

Systematic 5-Phase Methodology

1. Identify Failed Component - Find which PipelineRun/TaskRun failed
2. Log Analysis - Extract error messages from failed steps
3. Event Correlation - Match Kubernetes events to failure timeline
4. Resource Inspection - Examine YAML configurations for issues
5. Root Cause Analysis - Correlate findings to identify actual cause

Common Failure Patterns

Pre-built investigation workflows for:

• Image pull failures (wrong image, registry auth)
• Resource exhaustion (memory, CPU, quota)
• Build script failures (exit codes, missing tools)
• Timeout issues (slow operations, hangs)
• Workspace/volume problems (PVC, mounts)
• Permission errors (RBAC, ServiceAccount)

Standard Tool Usage

Uses kubectl and Tekton commands anyone can run:

kubectl get pipelinerun <name> -n <namespace>
kubectl logs <pod> -c step-<name> -n <namespace>
kubectl get events -n <namespace>
kubectl describe taskrun <name> -n <namespace>

No special tools or MCP servers required.

Example Usage

Basic Pipeline Failure

User: "My PipelineRun 'app-build-123' in namespace 'dev' failed. How do I debug it?"

Claude (with skill):
1. Check PipelineRun status: kubectl describe pipelinerun app-build-123 -n dev
2. Identify failed TaskRun from status
3. Get logs: kubectl logs <pod> -c step-<failed-step> -n dev
4. Check events: kubectl get events -n dev
5. Correlate to find root cause
6. Provide specific fix based on error found

ImagePullBackOff Error

User: "Getting ImagePullBackOff in my build pipeline."

Claude (with skill):
1. Describe pod to see events: kubectl describe pod <pod> -n <namespace>
2. Check exact error message in events
3. Verify image name and tag
4. Check ServiceAccount imagePullSecrets
5. Provide fix (correct image name, add secret, etc.)

Stuck Pipeline

User: "Pipeline running for 2 hours, should be done by now."

Claude (with skill):
1. List TaskRuns: kubectl get taskruns -l tekton.dev/pipelineRun=<name> -n <ns>
2. Identify which TaskRun is running/pending
3. Get logs from running TaskRun to see progress
4. Check if stuck (no new output) or slow (making progress)
5. Recommend timeout increase if legitimate, or investigate hang

What Makes This Skill Different

Systematic vs Ad-Hoc

• ✅ Follows repeatable 5-phase process
• ✅ Correlates multiple data sources
• ❌ Not random "try this" suggestions

Root Cause Focus

• ✅ Investigates before recommending fixes
• ✅ Distinguishes symptoms from causes
• ❌ Not quick fixes without understanding

General Purpose

• ✅ Uses standard kubectl/Tekton tools
• ✅ Works in any Konflux environment
• ❌ No special tools required

Skill Quality

• Format: Follows konflux-ci/skills conventions
• Testing: 6 comprehensive test scenarios
• TDD Approach: Built using Test-Driven Development for Documentation
• Word Count: ~950 words (under 1,000 technique skill target)
• Tool Agnostic: Uses standard Kubernetes/Tekton tools

Integration with Konflux Workflows

This skill complements:

• Konflux CI/CD pipeline development
• Build troubleshooting
• Integration testing
• On-call incident response
• Developer self-service debugging

Prerequisites

Required Knowledge

• Basic Tekton concepts (PipelineRun, TaskRun)
• kubectl command familiarity
• Understanding of container/pod concepts

Required Access

• kubectl access to Konflux namespace
• Permission to view PipelineRuns, TaskRuns, pods, events
• Access to pod logs

Optional

• Tekton CLI (tkn) for enhanced troubleshooting
• OpenShift CLI (oc) if using OpenShift-based Konflux

Common User Prompts

This skill works best with prompts like:

Good:

• "My PipelineRun 'component-build-xyz' in 'user-ns' failed"
• "Build stuck for 30 minutes, TaskRun 'build-step-abc'"
• "Getting ImagePullBackOff in pipeline 'deploy-prod'"
• "TaskRun shows OOMKilled, how do I fix it?"

Less Optimal:

• "Pipeline broken" (too vague)
• "Nothing works" (no specifics)
• "Help with Tekton" (not actionable)

Related Skills

Complements existing Konflux skills:

• understanding-konflux-resources - For resource basics
• Other pipeline or CI/CD skills in the repository

Contributing

Improvements welcome:

• Additional failure patterns
• Enhanced decision trees
• More kubectl examples
• Platform-specific tips (EKS, GKE, OpenShift)

License

Same as parent repository (Apache 2.0).

Authors

Gjorgji Georgievski (ggeorgie@redhat.com)

Systematic debugging methodology based on SRE best practices and real-world Konflux troubleshooting experience.

Version History

• 1.0.0 - Initial skill creation
  • 5-phase investigation methodology
  • 6 common failure patterns with kubectl commands
  • Root cause vs symptom distinction
  • Decision tree and troubleshooting workflow
  • Test-driven development with 6 scenarios