6hr job timeout doesn't appear to work as documented?

[snps-davidla]

Usage Limits Documentation:

• about-self-hosted-runners#usage-limits

Workflow Syntax Documentation:

• job timeout
• step timeout

Related Discussions/Issues:

• https://github.com/orgs/community/discussions/26679

This is not working when set at the job level alone.. We have a job called through a matrix in a reusable workflow and there appears to be no way to allow the job to run longer than 6hrs without setting the step timeout.

Our call stack looks like:

• toplevel repo workflow (3rd Party tool)
  • reusable workflow (multiplatform build/test/package)
  • matrix job (8 platforms)
    • reusable workflow (per platform build/test/package)
    • long running job (build/test/package)
      • long running step(s) (build or test)

I am trying:

• timeout at the long running step level
  • -> this appears to work?
• timeout at the long running job level within the deepest reusable workflow
  • -> no effect
• timeout at the matrix job level
  • -> timeout is not supported for reusable workflow jobs
• timeout at the repo toplevel reusable workflow job
  • -> timeout is not supported for reusable workflow jobs

Result: The long running jobs are consistently cancelled at 6hrs in all cases.