heartbeat.yml

# Missed-run heartbeat for the daily "Sync translation" cron.
#
# A failed sync run surfaces on its own, but a run that never fires is silent:
# GitHub can drop scheduled runs, auto-disable the workflow after repo inactivity,
# or the schedule can be removed. This workflow runs on its own schedule and
# alerts when the last *successful* scheduled sync is older than a threshold, so a
# silently missed run is caught. It does not depend on sync-translation.yml
# running, so it still fires once that workflow has stopped.

name: Sync heartbeat

on:
  schedule:
    # Noon UTC, offset from the sync's 00:00 run so a healthy day reads ~12h old.
    - cron: "0 12 * * *"
  workflow_dispatch:

permissions:
  contents: read
  actions: read

jobs:
  check-heartbeat:
    runs-on: ubuntu-latest
    steps:
      - name: Checkout repository (master)
        uses: actions/checkout@9c091bb21b7c1c1d1991bb908d89e4e9dddfe3e0 # v7.0.0
        with:
          persist-credentials: false

      - name: Check the last successful scheduled sync
        env:
          GH_TOKEN: ${{ secrets.GITHUB_TOKEN }}
        run: |
          set -euo pipefail

          # shellcheck source=assets/env.sh
          source "$GITHUB_WORKSPACE/.github/workflows/assets/env.sh"
          # shellcheck source=assets/lib.sh
          source "$GITHUB_WORKSPACE/.github/workflows/assets/lib.sh"

          begin_phase "$PHASE_HEARTBEAT" "Check daily sync heartbeat"
          trap '[[ -n "$CURRENT_PHASE" ]] && end_phase' EXIT ERR

          # createdAt is when the run was scheduled; updatedAt moves when an old
          # run is rerun, which would make a stopped schedule look healthy.
          last_ts="$(gh run list \
            --repo "$GITHUB_REPOSITORY" \
            --workflow sync-translation.yml \
            --event schedule \
            --status success \
            --limit 1 \
            --json createdAt \
            --jq '.[0].createdAt // ""')"

          last_epoch=""
          if [[ -n "$last_ts" ]]; then
            last_epoch="$(date -d "$last_ts" +%s)"
          fi
          now_epoch="$(date +%s)"
          max_age_seconds=$(( HEARTBEAT_MAX_AGE_HOURS * 3600 ))

          if heartbeat_run_is_stale "$last_epoch" "$now_epoch" "$max_age_seconds"; then
            # Interim alert: fail the job. Swap this for the shared notification
            # channel once the failure-notification work provides one.
            last_desc="${last_ts:-none on record}"
            echo "::error::Daily sync heartbeat: last successful scheduled run is stale (last=${last_desc}, threshold=${HEARTBEAT_MAX_AGE_HOURS}h). The scheduled sync may have stopped." >&2
            end_phase
            exit 1
          fi

          echo "Heartbeat healthy: last successful scheduled sync at ${last_ts} (within ${HEARTBEAT_MAX_AGE_HOURS}h)." >&2
          end_phase