Files
Charon/docs/implementation/ci_remediation_summary.md
akanealw eec8c28fb3
Some checks failed
Go Benchmark / Performance Regression Check (push) Has been cancelled
Cerberus Integration / Cerberus Security Stack Integration (push) Has been cancelled
Upload Coverage to Codecov / Backend Codecov Upload (push) Has been cancelled
Upload Coverage to Codecov / Frontend Codecov Upload (push) Has been cancelled
CodeQL - Analyze / CodeQL analysis (go) (push) Has been cancelled
CodeQL - Analyze / CodeQL analysis (javascript-typescript) (push) Has been cancelled
CrowdSec Integration / CrowdSec Bouncer Integration (push) Has been cancelled
Docker Build, Publish & Test / build-and-push (push) Has been cancelled
Quality Checks / Auth Route Protection Contract (push) Has been cancelled
Quality Checks / Codecov Trigger/Comment Parity Guard (push) Has been cancelled
Quality Checks / Backend (Go) (push) Has been cancelled
Quality Checks / Frontend (React) (push) Has been cancelled
Rate Limit integration / Rate Limiting Integration (push) Has been cancelled
Security Scan (PR) / Trivy Binary Scan (push) Has been cancelled
Supply Chain Verification (PR) / Verify Supply Chain (push) Has been cancelled
WAF integration / Coraza WAF Integration (push) Has been cancelled
Docker Build, Publish & Test / Security Scan PR Image (push) Has been cancelled
Repo Health Check / Repo health (push) Has been cancelled
History Rewrite Dry-Run / Dry-run preview for history rewrite (push) Has been cancelled
Prune Renovate Branches / prune (push) Has been cancelled
Renovate / renovate (push) Has been cancelled
Nightly Build & Package / sync-development-to-nightly (push) Has been cancelled
Nightly Build & Package / Trigger Nightly Validation Workflows (push) Has been cancelled
Nightly Build & Package / build-and-push-nightly (push) Has been cancelled
Nightly Build & Package / test-nightly-image (push) Has been cancelled
Nightly Build & Package / verify-nightly-supply-chain (push) Has been cancelled
Update GeoLite2 Checksum / update-checksum (push) Has been cancelled
Container Registry Prune / prune-ghcr (push) Has been cancelled
Container Registry Prune / prune-dockerhub (push) Has been cancelled
Container Registry Prune / summarize (push) Has been cancelled
changed perms
2026-04-22 18:19:14 +00:00

1.9 KiB
Executable File

CI Remediation Summary

Date: February 5, 2026 Task: Stabilize E2E testing pipeline and fix workflow timeouts.

Problem

The end-to-end (E2E) testing pipeline was experiencing significant instability, characterized by:

  1. Workflow Timeouts: Shard 4 was consistently timing out (>20 minutes), obstructing the CI process.
  2. Missing Dependencies: Security jobs for Firefox and WebKit were failing because they lacked the required Chromium dependency.
  3. Flaky Tests:
    • certificates.spec.ts failed intermittently due to race conditions when ensuring either an empty state or a table was visible.
    • crowdsec-import.spec.ts failed due to transient locks on the backend API.

Solution

Workflow Optimization

  • Shard Rebalancing: Reduced the number of shards from 4 to 3. This seemingly counter-intuitive move rebalanced the test load, preventing the specific bottlenecks that were causing Shard 4 to hang.
  • Dependency Fix: Explicitly added the Chromium installation step to Firefox and WebKit security jobs to ensure all shared test utilities function correctly.

Test Logic Improvements

  • Robust Empty State Detection: Replaced fragile boolean checks with Playwright's .or() locator pattern.
    • Old: isVisible().catch() (Bypassed auto-waits, led to race conditions)
    • New: expect(locatorA.or(locatorB)).toBeVisible() (Leverages built-in retry logic)
  • Resilient API Retries: Implemented .toPass() for the CrowdSec import test.
    • This allows the test to automatically retry the import request with exponential backoff if the backend is temporarily locked or busy, significantly reducing flakes.

Results

  • Stability: The "Empty State OR Table" flake in certificates is resolved.
  • Reliability: CrowdSec import tests now handle transient backend states gracefully.
  • Performance: CI jobs now complete within the allocated time budget with balanced shards.