LookML CI/CD Pipeline: Complete Analysis

Pipeline Overview

CI/CD pipeline is a comprehensive LookML Validation Pipeline that automatically tests and validates LookML code changes. It runs on GitHub Actions and performs 5 different types of validation to ensure code quality, functionality, and reliability.

Trigger Events

The pipeline runs on:

• Pull Requests to main, master, or develop branches
• Direct pushes to main or master branches
• Only when LookML-related files are changed (.lkml, .lookml, .view, etc.)

File Structure & Responsibilities

Main Workflow File

• .github/workflows/lookml-validation.yml: The orchestrator that defines the entire pipeline flow

Validation Scripts (Python)

• looker-validator.py: LookML syntax validation using Looker API
• lookml-linter.py: Custom coding standards and best practices checker
• sql-execution-validator.py: Tests actual SQL execution against databases
• lookml-data-tests.py: Runs LookML data tests and model validation
• lookml-content-validator.py: Validates Looker content (dashboards, looks)

Utility Scripts

• report-generator.py: Creates comprehensive validation reports
• check-results.py: Determines overall pipeline success/failure

Configuration Files

• linting-rules.yaml: Defines coding standards and validation rules
• requirements.txt: Python package dependencies
• looker-config.ini.template: Template for Looker API configuration

Pipeline Steps Breakdown

Step 1: Environment Setup

• Checkout code (fetch-depth: 0 for full history)
• Setup Python 3.9
• Cache pip dependencies for performance
• Install Python packages from requirements.txt
• Configure Looker API credentials from GitHub secrets

Step 2: Change Detection

• Uses tj-actions/changed-files@v44 to detect modified LookML files
• If no LookML files changed, pipeline skips validation steps
• Displays list of changed files for transparency

Step 3: LookML Syntax Validation

Script: looker-validator.py

• Purpose: Validates LookML syntax using official Looker API
• Process:
  • Authenticates with Looker API
  • Runs project validation for bi_sandbox
  • Handles both 200 (with errors) and 204 (success) responses
  • Falls back to manifest.lkml validation if project not found
• Output: validation_results.json
• Blocking: ❌ YES (pipeline fails if syntax errors found)

Step 4: Custom LookML Linting

Script: lookml-linter.py

• Purpose: Enforces coding standards and best practices
• Checks:
  • Mandatory fields: Ensures dimensions/measures have required properties (label, description, type)
  • Naming conventions: Enforces snake_case, boolean prefixes (is_, has_)
  • Performance rules: Detects SELECT *, excessive nesting
  • Security rules: Identifies potential hardcoded secrets, PII without access filters
• Configuration: Rules defined in linting-rules.yaml
• Output: linting_results.json
• Blocking: ❌ YES (errors fail pipeline, warnings don't)

Step 5: SQL Execution Testing

Script: sql-execution-validator.py

• Purpose: Tests that SQL actually executes against real databases
• Process:
  • Extracts SQL from derived tables in LookML files
  • Uses 2-step Looker API workflow:
    • Create SQL query (POST /api/4.0/sql_queries)
    • Execute query (POST /api/4.0/sql_queries/{slug}/run/json)
  • Detects database connections from model files
  • Tests SQL syntax, table existence, permissions
• Advanced Features:
  • Handles multiline SQL formatting
  • Detects connection from model files or derived_table
  • Parses execution results for hidden errors (even with 200 status)
• Output: sql_validation_results.json
• Blocking: ❌ YES (SQL execution errors fail pipeline)

Step 6: LookML Data Tests

Script: lookml-data-tests.py

• Purpose: Validates LookML tests and model integrity
• Process:
  • Runs model validation for all models in project
  • Validates explore definitions
  • Checks LookML test definitions (like in case21_tests.lkml)
  • Verifies model-explore relationships
• Tests Your Cases:
  • org_value_value_check: Tests value > 0
  • org_value_dep_length_check: Tests department length = 5
  • org_value_depid_notnull_check: Tests NOT NULL constraints
  • dep_a_good_number_limit: Tests total_number < 10000
• Output: data_tests_results.json
• Blocking: ⚠️ PARTIAL (continues on error, but affects final status)

Step 7: Content Validation

Script: lookml-content-validator.py

• Purpose: Validates Looker content (dashboards, looks) in BI Sandbox folder
• Process:
  • Uses Looker's content validation API
  • Filters results for "BI Sandbox" folder only
  • Checks dashboard elements, scheduled plans, alerts
  • Validates content dependencies and model references
• Output: content_validation_results.json
• Blocking: ⚠️ NO (allows pipeline to continue with warnings)

Step 8: Report Generation 📋

Script: report-generator.py

• Purpose: Creates comprehensive validation reports
  • Generates:
  • GitHub Actions step summary (visible in workflow)
  • PR comment with detailed results and status icons
  • Markdown summary file
• Features:
  • Status icons (✅ ❌ ⚪) for each validation step
  • Error details with file paths and line numbers
  • Performance metrics (execution times, query counts)

Step 9: Artifact Upload & Results Check

• Uploads: All JSON results and summary files to GitHub
• Retention: 30 days for debugging and history
• Final Check: check-results.py determines overall success/failure

Step 10: PR Commenting 💬

• Updates existing PR comments or creates new ones
• Shows validation status with visual indicators
• Provides direct links to detailed workflow results

Output Artifacts

Each validation step produces detailed JSON results:

1. validation_results.json - LookML syntax validation
2. linting_results.json - Coding standards results
3. sql_validation_results.json - SQL execution test results
4. data_tests_results.json - LookML data tests results
5. content_validation_results.json - Content validation results
6. validation_summary.md - Human-readable summary
7. pr_comment.md - PR comment content

Error Handling & Resilience

Blocking vs Non-Blocking Steps

• Blocking (Pipeline Fails): Syntax errors, linting errors, SQL execution errors
• Non-Blocking (Warnings Only): Content validation issues, some data test failures

Error Categories

• Syntax Errors: Malformed LookML, missing required fields
• Execution Errors: SQL that fails to run, missing tables/permissions
• Content Errors: Broken dashboards, invalid model references
• Security Errors: Hardcoded secrets, unprotected PII fields