Timeouts & diagnosable failure

Bound the cost of a stuck job

Lesson 19 of 20 · 12 min · Module 6: Verify and recover

A job that never finishes can block feedback and consume runner capacity.

Start this lesson →

Free. No installation: you write and run the code in your browser, and it is checked automatically.

By the end, you’ll be able to…

  • Give failure a time boundary
  • Record useful diagnostic evidence

The idea

A workflow combines an event boundary, job dependencies, and ordered steps. Conditions and prerequisite outcomes determine which jobs can run.

01Give failure a time boundary

A timeout limits how long the workflow waits for an unresponsive command. The appropriate value should reflect expected duration and variability.

02Record useful diagnostic evidence

A timeout message should identify the stuck responsibility. Logs, timestamps, and the evaluated source revision make the next investigation more focused.

03Avoid replacing diagnosis with repeated retries

A retry can be appropriate for transient failures, but repeated unbounded retries can hide a deterministic fault and increase cost.

Your turn

Define test with timeout-minutes 10, runs-on ubuntu-latest, and npm test. The model validates the timeout declaration; the real Actions scheduler enforces it.

What the checks look for

  • Declares a ten-minute job limit
  • Keeps a real check step
Try it in the editor →

Take this with you

Bound work, preserve evidence, and investigate repeated failures.

Go a little further

Define a useful retry policy for a transient registry download failure.

Learn CI/CD & Release Engineering for free

20 short lessons with real code and instant checks.

See the whole course →