Auto-rerun failed CI jobs #25

	---

	name: Auto-rerun failed CI jobs

	# See PR #659: exit-143 SIGTERM crashes in MPI-using jobs correlate with
	# specific Azure regions (centralus/northcentralus/westcentralus/westus3),
	# not with any package version or OS image, and persist for the life of a
	# runner - retrying a command in-place doesn't help when the problem is
	# the runner itself (confirmed: two in-place retry attempts in the same
	# job both failed identically). A fresh job attempt gets assigned a new
	# runner and a new shot at a healthy region.
	#
	# This has to be a separate workflow_run-triggered workflow rather than a
	# job inside ci_pipeline.yml: GitHub refuses `gh run rerun` while the
	# target run still has an active job ("This workflow is already running"),
	# and a job calling it from within its own run can never observe that run
	# as anything but in-progress. workflow_run only fires once the run has
	# genuinely finished.

	on:
	workflow_run:
	workflows: ["CI pipeline for pySDC"]
	types: ["completed"]

	jobs:
	rerun-failed-jobs:
	runs-on: ubuntu-latest
	if: >-
	${{ github.repository_owner == 'Parallel-in-Time'
	&& github.event.workflow_run.conclusion == 'failure'
	&& github.event.workflow_run.run_attempt < 3 }}
	permissions:
	actions: write
	steps:
	- name: Rerun failed jobs in a new attempt
	env:
	GH_TOKEN: ${{ github.token }}
	run: \|
	gh run rerun ${{ github.event.workflow_run.id }} --repo ${{ github.repository }} --failed

Navigation Menu