The One-Liner

Cancel older running pipelines/workflows on the same branch (keep latest)

Hey engineer, happy Friday! 🎉

Today we’re taking a look at a one-liner power move for CI/CD that automatically cancels older duplicate runs for a branch and keeps the latest which lowers CI costs and gives you faster feedback.

I’ll give you GitHub Actions, GitLab, and a gh (GitHub CLI) variant. Each has a dry-run option and safe caveats.

---

1. GitHub Actions (curl + jq) — cancel older runs for branch-name

# env: GITHUB_TOKEN, OWNER, REPO, BRANCH
curl -s -H “Authorization: Bearer $GITHUB_TOKEN” \
  “https://api.github.com/repos/$OWNER/$REPO/actions/runs?per_page=100” \
  | jq -r --arg BR “$BRANCH” ‘.workflow_runs 
      | map(select(.head_branch==$BR and (.status==”in_progress” or .status==”queued”)))
      | sort_by(.created_at) 
      | .[0:-1][] | .id’ \
  | xargs -r -n1 -I{} curl -s -X POST -H “Authorization: Bearer $GITHUB_TOKEN” \
      “https://api.github.com/repos/$OWNER/$REPO/actions/runs/{}/cancel”

Dry-run: replace the final curl -X POST ... with echo “cancel {}” inside the pipeline (or remove the | xargs ...and print IDs/URLs).

What it does: lists running/queued workflow runs for the branch, sorts by creation time, and cancels all but the newest.

---

2. GitHub CLI (cleaner) — same behaviour with gh

# env: GH_TOKEN set & gh authenticated, OWNER/REPO set or run within repo
gh api repos/$OWNER/$REPO/actions/runs --paginate --jq \
  --jq ‘.workflow_runs | map(select(.head_branch==env.BRANCH and (.status==”in_progress” or .status==”queued”))) | sort_by(.created_at) | .[0:-1][] | .id’ \
  | xargs -r -n1 -I{} gh api -X POST repos/$OWNER/$REPO/actions/runs/{}/cancel

Note: gh simplifies auth and pagination; use --jq/--paginate carefully.

---

3. GitLab — cancel older pipelines for a branch, keep the most recent

# env: GITLAB_TOKEN, PROJECT_ID, BRANCH
curl -s --header “PRIVATE-TOKEN: $GITLAB_TOKEN” \
  “https://gitlab.com/api/v4/projects/$PROJECT_ID/pipelines?ref=$BRANCH&per_page=100” \
  | jq -r ‘sort_by(.id) | .[0:-1][] | .id’ \
  | xargs -r -n1 -I{} curl -s -X POST --header “PRIVATE-TOKEN: $GITLAB_TOKEN” \
      “https://gitlab.com/api/v4/projects/’$PROJECT_ID’/pipelines/{}/cancel”

Dry-run: print pipeline IDs/URLs instead of calling cancel.

---

Quick breakdown & tips

• sort_by(.created_at) (GitHub) or sort_by(.id) (GitLab) orders runs; removing the last (.[0:-1]) cancels all but the most recent.

• xargs -r avoids errors when no IDs are returned.

• Use per_page / pagination (or --paginate with gh) if you have many runs.

• Run this in a pre-commit hook, a scheduled job, or as a short-lived GitHub Action/GitLab job triggered on push. Good to run as a small maintenance job for busy repos.

• Dry-run first. Always list IDs/URLs before canceling to make sure you won’t cancel a deployment or other critical run.

---

Caveats & best practices

• Auth & scopes: Token must have permissions (GitHub: actions write; GitLab: api/maintainer as required). Don’t hardcode tokens. Use repo secrets or a secrets manager.

• Deploy safety: Some pipelines include deploy steps; prefer canceling only test/build pipelines or ensure your pipeline is safe to be canceled mid-run (e.g., deploy steps gated or flagged).

• Rate limits & pagination: If you query a very busy repo, use pagination and throttle requests to avoid API limits.

• Run selection: Consider narrowing by workflow_id, workflow_name, or actor to avoid canceling unrelated runs.

• Idempotency: Cancelling a run that’s about to finish wastes a tiny amount of time; acceptable tradeoff vs. queue buildup. If unsure, add minimal age filter (created_at < now - 30s) to avoid racing the latest run.

---

Variations you’ll love

• Only cancel queued runs (leave in_progress): change status filter to only queued.

• Cancel older runs across all branches except main: don’t filter by branch, but exclude main or protected branches.

• Auto-trigger after cancel (re-run latest): after canceling older runs you can explicitly re-run the latest to ensure a clean queue.

---

Until next time,
Stay sharp.

---

For those that want speed, reliability and best practices without starting from scratch, I released CI/CD Pipeline Templates — a collection of fully functional pipelines for GitHub Actions, GitLab CI, Jenkins, and CircleCI.

---

Don’t waste another weekend writing Bash scripts from scratch.

The DevOps Automation Pack gives you 60+ plug-and-play production-ready Bash scripts for backups, monitoring, security, cleanup, and more.

Every script is cron-ready, supports dry-run mode, and works right out of the box.

---