LogoLogo
SlackLogin
  • Overview
    • Welcome
  • Setup & Configuration
    • Connecting to Trunk
    • Managing Your Organization
      • GitHub App Permissions
  • Integration with Slack
  • Flaky Tests (Beta)
    • Overview
    • Get Started
      • Test Frameworks
        • Android
        • Bazel
        • Behave
        • cargo-nextest
        • Cypress
        • Dart Test
        • Go
        • GoogleTest
        • Gradle
        • Jasmine
        • Jest
        • Karma
        • Kotest
        • Maven
        • minitest
        • Mocha
        • Nightwatch
        • NUnit
        • Pest
        • PHPUnit
        • Playwright
        • Pytest
        • Robot Framework
        • RSpec
          • RSpec (Manual Uploads)
        • Swift Testing
        • Vitest
        • XCTest
        • Other Test Frameworks
      • CI Providers
        • Azure DevOps Pipelines
        • BitBucket Pipelines
        • Buildkite
        • CircleCI
        • Drone CI
        • GitHub Actions
        • GitLab
        • Jenkins
        • Semaphore CI
        • TeamCity
        • Travis CI
        • Other CI Providers
    • Dashboard
    • Flaky Test Detection
    • Quarantining
    • PR Comments
    • Ticketing Integrations
      • Jira Integration
      • Linear Integration
      • Other Ticketing Platforms
    • Webhooks
      • Slack Integration
      • Microsoft Teams Integration
      • GitHub Issues Integration
      • Linear Integration
    • Uploader CLI Reference
  • Merge Queue
    • Overview
    • How does it work?
    • Setup
      • Quick Start
      • Settings
      • Integration for Slack
    • Concepts and Optimizations
      • Predictive Testing
      • Optimistic Merging
      • Pending Failure Depth
      • Anti-Flake Protection
      • Batching
      • Parallel Queues
        • Bazel
        • Nx
        • API
      • FAQ
    • Priority
    • Managing Merge Queue
      • Using the Merge UI
      • Metrics
      • Command Line
    • Webhooks
    • Reference
  • Code Quality
    • Overview
    • Why Metalinters?
      • How does it work?
      • Why Code Quality?
    • Setup & Installation
      • Initialize Trunk
      • Local Linting
      • Linting in CI
      • Nightly Report (Deprecated)
      • OpenAI Codex Support
    • IDE Integration
      • VSCode
      • Neovim
      • GitHub Codespaces
    • Linters
      • Supported Linters
        • Actionlint
        • Ansible-lint
        • Autopep8
        • Bandit
        • Biome
        • Black
        • Brakeman
        • buf
        • Buildifier
        • cfnlint
        • Checkov
        • circleci
        • ClangFormat
        • clang-tidy
        • Clippy
        • cmake-format
        • codespell
        • cspell
        • cue-fmt
        • dart
        • deno
        • Detekt
        • djlint
        • dotenv-linter
        • dotnet-format
        • dustilock
        • ESLint
        • Flake8
        • git-diff-check
        • Gitleaks
        • Gofmt
        • gofumpt
        • goimports
        • gokart
        • golangci-lint
        • golines
        • google-java-format
        • graphql-schema-linter
        • hadolint
        • haml-lint
        • isort
        • iwyu
        • ktlint
        • kube-linter
        • markdown-link-check
        • markdown-table-prettify
        • Markdownlint
        • markdownlint-cli2
        • mypy
        • nancy
        • nixpkgs-fmt
        • opa
        • OSV-Scanner
        • Oxipng
        • perlcritic
        • perltidy
        • php-cs-fixer
        • phpstan
        • pmd
        • pragma-once
        • pre-commit-hooks
        • Prettier
        • prisma
        • psscriptanalyzer
        • Pylint
        • pyright
        • regal
        • remark-lint
        • renovate
        • rome
        • rubocop
        • Ruff
        • rufo
        • rustfmt
        • scalafmt
        • semgrep
        • ShellCheck
        • shfmt
        • sort-package-json
        • sourcery
        • sql-formatter
        • SQLFluff
        • sqlfmt
        • squawk
        • standardrb
        • stringslint
        • stylelint
        • stylua
        • SVGO
        • swiftformat
        • swiftlint
        • taplo
        • Terraform
        • terragrunt
        • terrascan
        • TFLint
        • tfsec
        • tofu
        • Trivy
        • Trufflehog
        • txtpbfmt
        • vale
        • Yamllint
        • yapf
      • Run Linters
      • Manage Linters
      • Configure Linters
      • Ignoring Issues and Files
      • Custom Linters
      • Shared Configs
      • Upgrades
    • Debugging
    • Licensing
  • CLI & API References
    • CLI Reference
      • Install
      • Getting Started
        • Code Quality
        • Merge Queue
        • Flaky Tests
        • Tools
        • Actions
          • Git Hooks
        • Announce
      • Compatibility
      • Caching
      • Commands Reference
        • Code Quality
        • Actions
        • Merge
      • Configuration
        • Plugins
          • Share Config Between Codebases
          • Exporting Linter Configs
        • Runtimes
        • Tools
        • Lint
          • Definitions
          • Commands
          • Output
          • Output Parsing
          • Files and Caching
          • Dependencies
          • Auto-Enable
        • Actions
          • Notifications
          • Logging and Troubleshooting
        • Merge
        • Telemetry
        • Per User Overrides
    • API Reference
      • Flaky Tests
      • Merge Queue
      • Webhooks Reference
  • Pricing & Security
    • Security
  • Billing
  • Community & Support
  • Links
    • Open App
    • Slack Community
    • Changelog
    • Feature Requests
On this page
Edit on GitHub
  1. Merge Queue
  2. Concepts and Optimizations

Pending Failure Depth

Last updated 2 months ago

By default, a PR that fails testing will be evicted from the queue. The Pending Failure Depth feature allows a failed PR to remain in the queue for pull requests behind it so that testing can be finished before this eviction occurs. The number of PRs that the queue will wait for is the Pending Failure Depth. This depth is configurable and reflects the number of pull requests behind this one that should complete testing before eviction is assessed. Let's take the example queue below: main <- A <- B <- C If pending failure depth is set to 2 and A fails, Trunk Merge Queue will wait until both B and C complete testing before retrying or kicking A out of the queue. Why does this matter? Since Trunk Merge Queue performs predictive testing when the queue above is tested, it is actually testing the code like this: main <- A <- B+a <- C+ba The code in A is actually being tested when both B and C are being tested. We can take advantage of this feature of to give pull request A a second and third chance to pass tests. That means that if pull request A fails because of a flaky test, it doesn't have to be immediately rejected and can leverage the test coverage it receives during the testing of B and C. By combining pending failure depth with optimistic merging - we can create a merge queue better .

predictive testing
protected from flaky failures