Optimize Pipeline

Instructions

Step 1: Establish Scope

Call MCP tool: harness_list
Parameters:
  resource_type: "pipeline"
  org_id: "<organization>"
  project_id: "<project>"

Call MCP tool: harness_list
Parameters:
  resource_type: "execution"
  org_id: "<organization>"
  project_id: "<project>"
  pipeline_id: "<pipeline_identifier>"

Step 2: Identify the Optimization Task

1. Parallel Testing with Test Intelligence -- Split tests across runners and skip unchanged tests
2. Caching Strategy -- Multi-layer dependency, build output, and test result caching
3. Pipeline Bottleneck Analysis -- Stage-level timing breakdown with recommendations
4. Cache Hit Rate Improvement -- Diagnose and fix low cache hit rates
5. Monorepo CI Pipeline -- Selective builds triggered by changed paths

Step 3: Configure Parallel Testing with Test Intelligence

• Test framework (JUnit, pytest, Jest, Go test, etc.)
• Total test count and current runtime
• Target runtime

• Split tests across N parallel runners using Harness Test Intelligence
• Use TI to identify and skip unchanged tests based on code changes
• Configure test splitting method: by class, by file, or by timing data
• Set up test result aggregation across runners
• Track TI savings over time (tests skipped vs. total)

• Enable Test Intelligence in the pipeline stage
• Set parallelism level based on test count and runner capacity
• Configure test report collection from all parallel runners
• Set up failure thresholds (e.g., fail the stage if any runner fails)

Step 4: Design Caching Strategy

• Build tool (Maven, Gradle, npm, yarn, pip, Go modules)
• Current build time breakdown (dependency download, compile, test)
• Cache key source (lockfile hash, manifest hash)

• Cache key: hash of lockfile (package-lock.json, pom.xml, go.sum)
• Cache path: dependency directory (~/.m2, node_modules, ~/.cache/pip)
• Fallback: use previous cache if exact match not found

• Cache key: hash of source files
• Cache path: build output directory (target/, dist/, build/)
• Invalidation: any source file change

• Cache key: hash of source + test files
• Cache path: test result and coverage directories
• Use to skip unchanged tests in combination with TI

Step 5: Analyze Pipeline Bottlenecks

Call MCP tool: harness_get
Parameters:
  resource_type: "execution"
  resource_id: "<execution_id>"
  org_id: "<organization>"
  project_id: "<project>"

• Duration vs. pipeline total (find the longest stages)
• Queue time vs. execution time (runner availability issues)
• Sequential stages that could run in parallel
• Steps that download large artifacts repeatedly

Step 6: Design Monorepo CI Pipeline

• Number of services and their directories
• Shared libraries and their dependents
• Build tool

• Use path-based triggers: only build services with changed files
• Build shared libraries when they change, then rebuild all dependents
• Run integration tests only when cross-service dependencies change
• Use a dependency graph to determine the minimal build set

Examples

• "My pipeline takes 45 minutes, help me speed it up" -- Analyze bottlenecks and recommend parallel testing, caching, and stage reordering
• "Set up parallel tests with Test Intelligence" -- Configure TI with test splitting across N runners
• "Improve our cache hit rate" -- Diagnose cache key configuration and fix common misses
• "Design a CI pipeline for our monorepo with 5 services" -- Configure path-based triggers with selective builds
• "Our builds download dependencies every time" -- Design multi-layer caching strategy with fallback

Performance Notes

• Test Intelligence needs 2-3 full test runs to build its initial model -- first runs will execute all tests.
• Cache keys should be based on lockfiles, not timestamps -- timestamps cause unnecessary cache misses.
• Parallelism beyond runner capacity causes queuing -- profile available runner capacity before increasing parallelism.
• Monorepo path triggers should include shared library directories to avoid missing transitive changes.

Troubleshooting

Test Intelligence Not Skipping Tests

• Verify TI is enabled and has completed baseline runs
• Check that the test framework is supported by Harness TI
• Ensure test report format is correctly configured for the framework

Cache Misses on Every Build

• Check cache key configuration -- keys should use file hashes, not timestamps
• Verify cache path matches the actual dependency directory
• Check that the cache storage backend is accessible from all runners

Monorepo Building Everything

• Verify path-based triggers are configured correctly
• Check that the dependency graph includes shared library paths
• Ensure glob patterns in triggers match the actual directory structure