Skip to content

feat: adopt fleet reconcile gate and fix preview pin #70

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
joshua-temple merged 1 commit into from
Jun 24, 2026
Merged

feat: adopt fleet reconcile gate and fix preview pin #70

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
4 changes: 2 additions & 2 deletions .github/workflows/cascade-pr-preview.yaml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -24,9 +24,9 @@ jobs:
fetch-depth: 0

- name: Setup CLI
uses: stablekernel/cascade/.github/actions/setup-cli@v0.3.0-rc.26
uses: stablekernel/cascade/.github/actions/setup-cli@v0.2.1
with:
version: v0.3.0-rc.26
version: v0.2.1

- name: Validate manifest
run: |
Expand Down
175 changes: 175 additions & 0 deletions .github/workflows/scenario-suite.yaml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -18,7 +18,24 @@ jobs:
reset:
name: Reset manifest state
runs-on: ubuntu-latest
outputs:
# Captured before anything is dispatched or merged. The reconcile job
# enumerates every run this repo produced at or after this instant and
# fails if any is unaccounted for in the ledger. Recorded first, in the
# earliest job, so no run the suite causes can fall outside the window.
window-start: ${{ steps.window.outputs.window-start }}
steps:
# The reconcile window opens here, before the first reset push, dispatch,
# or merge. Every run the suite goes on to cause lands at or after this
# timestamp, so the reconcile job sees all of them.
- name: Open reconcile window
id: window
run: |
set -euo pipefail
WINDOW_START="$(date -u +%Y-%m-%dT%H:%M:%SZ)"
echo "window-start=$WINDOW_START" >> "$GITHUB_OUTPUT"
echo "reconcile window opened at $WINDOW_START"

- name: Checkout
uses: actions/checkout@v4
with:
Expand Down Expand Up @@ -119,6 +136,12 @@ jobs:
git commit --no-gpg-sign -m "feat: dummy src change for scenario test"
git push origin "$BRANCH"

# The seed branch head SHA keys the Cascade PR Preview run this PR
# fires. Capture it before the squash-merge deletes the branch so the
# reconcile ledger can register that preview run by head SHA.
SEED_HEAD_SHA="$(git rev-parse HEAD)"
echo "seed_head_sha=$SEED_HEAD_SHA" >> "$GITHUB_OUTPUT"

gh pr create --base main --head "$BRANCH" \
--title "feat: dummy src change for scenario test" \
--body "Automated scenario run; drives orchestrate on merge."
Expand Down Expand Up @@ -170,6 +193,52 @@ jobs:

echo "run_id=$RUN_ID" >> "$GITHUB_OUTPUT"

- name: Resolve seed PR Cascade PR Preview run
id: seed_preview
env:
GH_TOKEN: ${{ secrets.CASCADE_STATE_TOKEN }}
run: |
set -euo pipefail
# The seed PR fired the Cascade PR Preview workflow on its head SHA.
# Historically the suite never read this run's conclusion, so a broken
# generated preview workflow (e.g. a stale setup-cli pin) stayed
# invisible. Resolve its run id here so the reconcile ledger registers
# it expected:success; from now on a red seed preview reds the suite.
SEED_HEAD_SHA="${{ steps.seed.outputs.seed_head_sha }}"
MAX_ATTEMPTS=5
ATTEMPT=0
RUN_ID=""
while [ "$ATTEMPT" -lt "$MAX_ATTEMPTS" ]; do
RUN_ID=$(gh run list \
--workflow=cascade-pr-preview.yaml \
--json=databaseId,headSha \
--jq=".[] | select(.headSha==\"$SEED_HEAD_SHA\") | .databaseId" 2>/dev/null | head -n1 || echo "")
[ -n "$RUN_ID" ] && break
ATTEMPT=$((ATTEMPT + 1))
sleep 60
done
if [ -z "$RUN_ID" ]; then
echo "::error::Cascade PR Preview run for seed head $SEED_HEAD_SHA did not appear within timeout"
exit 1
fi
echo "run_id=$RUN_ID" >> "$GITHUB_OUTPUT"

- name: Register seed PR Cascade PR Preview run
uses: stablekernel/cascade/.github/actions/register-run@main
with:
run-id: ${{ steps.seed_preview.outputs.run_id }}
expected-conclusion: success
reason: seed-pr-preview
upload: 'true'

- name: Register seed orchestrate run
uses: stablekernel/cascade/.github/actions/register-run@main
with:
run-id: ${{ steps.wait_orchestrate.outputs.run_id }}
expected-conclusion: success
reason: seed-orchestrate-merge
upload: 'true'

- name: Assert per-leg matrix artifacts exist (artifacts API)
env:
GH_TOKEN: ${{ secrets.GITHUB_TOKEN }}
Expand Down Expand Up @@ -368,6 +437,14 @@ jobs:

echo "run_id=$RUN_ID" >> "$GITHUB_OUTPUT"

- name: Register promote run
uses: stablekernel/cascade/.github/actions/register-run@main
with:
run-id: ${{ steps.wait_promote.outputs.run_id }}
expected-conclusion: success
reason: promote-staging-to-prod
upload: 'true'

- name: Assert manifest state shape (release marker populated)
run: |
# A terminal promotion publishes the release and records the version
Expand Down Expand Up @@ -475,6 +552,11 @@ jobs:
git add src/rollback-marker.txt
git commit --no-gpg-sign -m "feat: second staging version for rollback test"
git push origin "$BRANCH"
# Capture the seed branch head SHA before the squash-merge deletes the
# branch; it keys the Cascade PR Preview run this PR fires, which the
# reconcile ledger registers below.
SEED_HEAD_SHA="$(git rev-parse HEAD)"
echo "seed_head_sha=$SEED_HEAD_SHA" >> "$GITHUB_OUTPUT"
gh pr create --base main --head "$BRANCH" \
--title "feat: second staging version for rollback test" \
--body "Automated scenario run; drives orchestrate on merge."
Expand All @@ -484,6 +566,7 @@ jobs:
echo "merge_sha=$MERGE_SHA" >> "$GITHUB_OUTPUT"

- name: Wait for orchestrate to deploy the second staging version
id: wait_second_orchestrate
env:
GH_TOKEN: ${{ secrets.CASCADE_STATE_TOKEN }}
run: |
Expand Down Expand Up @@ -520,6 +603,51 @@ jobs:
ATTEMPT=$((ATTEMPT + 1))
sleep 60
done
echo "run_id=$RUN_ID" >> "$GITHUB_OUTPUT"

- name: Resolve second seed PR Cascade PR Preview run
id: rollback_seed_preview
env:
GH_TOKEN: ${{ secrets.CASCADE_STATE_TOKEN }}
run: |
set -euo pipefail
# The rollback seed PR also fires the Cascade PR Preview workflow on its
# head SHA. Resolve its run id so reconcile registers it expected:success
# the same way as the first seed PR.
SEED_HEAD_SHA="${{ steps.second.outputs.seed_head_sha }}"
MAX_ATTEMPTS=5
ATTEMPT=0
RUN_ID=""
while [ "$ATTEMPT" -lt "$MAX_ATTEMPTS" ]; do
RUN_ID=$(gh run list \
--workflow=cascade-pr-preview.yaml \
--json=databaseId,headSha \
--jq=".[] | select(.headSha==\"$SEED_HEAD_SHA\") | .databaseId" 2>/dev/null | head -n1 || echo "")
[ -n "$RUN_ID" ] && break
ATTEMPT=$((ATTEMPT + 1))
sleep 60
done
if [ -z "$RUN_ID" ]; then
echo "::error::Cascade PR Preview run for rollback seed head $SEED_HEAD_SHA did not appear within timeout"
exit 1
fi
echo "run_id=$RUN_ID" >> "$GITHUB_OUTPUT"

- name: Register second seed PR Cascade PR Preview run
uses: stablekernel/cascade/.github/actions/register-run@main
with:
run-id: ${{ steps.rollback_seed_preview.outputs.run_id }}
expected-conclusion: success
reason: rollback-seed-pr-preview
upload: 'true'

- name: Register second seed orchestrate run
uses: stablekernel/cascade/.github/actions/register-run@main
with:
run-id: ${{ steps.wait_second_orchestrate.outputs.run_id }}
expected-conclusion: success
reason: rollback-seed-orchestrate
upload: 'true'

- name: Dispatch cascade-rollback.yaml for staging
id: dispatch_rollback
Expand Down Expand Up @@ -572,6 +700,14 @@ jobs:
gh run watch "$RUN_ID" --exit-status --interval 60
echo "run_id=$RUN_ID" >> "$GITHUB_OUTPUT"

- name: Register cascade-rollback run
uses: stablekernel/cascade/.github/actions/register-run@main
with:
run-id: ${{ steps.wait_rollback.outputs.run_id }}
expected-conclusion: success
reason: rollback-dispatch
upload: 'true'

- name: Assert staging rolled back to the prior version and marked diverged
run: |
set -euo pipefail
Expand Down Expand Up @@ -699,6 +835,14 @@ jobs:
gh run watch "$RUN_ID" --exit-status --interval 60
echo "run_id=$RUN_ID" >> "$GITHUB_OUTPUT"

- name: Register dispatched orchestrate run
uses: stablekernel/cascade/.github/actions/register-run@main
with:
run-id: ${{ steps.wait_dispatch.outputs.run_id }}
expected-conclusion: success
reason: dispatch-inputs-orchestrate
upload: 'true'

- name: Assert dispatched reason reached consolidated state
env:
GH_TOKEN: ${{ secrets.CASCADE_STATE_TOKEN }}
Expand Down Expand Up @@ -835,6 +979,14 @@ jobs:
gh run watch "$RUN_ID" --exit-status --interval 60
echo "run_id=$RUN_ID" >> "$GITHUB_OUTPUT"

- name: Register preview-check Cascade PR Preview run
uses: stablekernel/cascade/.github/actions/register-run@main
with:
run-id: ${{ steps.wait_preview.outputs.run_id }}
expected-conclusion: success
reason: preview-check-run
upload: 'true'

- name: Assert preview comment was posted (hard fail)
env:
GH_TOKEN: ${{ secrets.GITHUB_TOKEN }}
Expand Down Expand Up @@ -890,3 +1042,26 @@ jobs:
echo "| cascade-pr-preview run succeeded | ✓ |"
echo "| Preview comment posted by github-actions[bot] (marker) | ✓ |"
} >> "$GITHUB_STEP_SUMMARY"

# ---------------------------------------------------------------------
# Reconcile gate: structural coverage backstop. The per-job asserts above
# still gate each run they remember to wait on; this job is additive. It
# enumerates EVERY run this repo produced since window-start (captured in the
# reset job before the first push, dispatch, or merge) and fails if any is
# unaccounted for in the ledger the register-run steps uploaded - an
# unregistered non-success run, or a registered run that concluded other than
# its expected conclusion. That turns any fire-and-forget run the suite forgot
# to gate (historically the seed PR's Cascade PR Preview run) into a hard red.
reconcile:
name: Reconcile scenario-window runs
needs: [reset, commit-and-build, promote-staging, rollback-check, dispatch-inputs-check, pr-preview-check]
if: always()
uses: stablekernel/cascade/.github/workflows/fleet-reconcile.yaml@main
permissions:
contents: read
actions: read
with:
window-start: ${{ needs.reset.outputs.window-start }}
# Artifact mode: each register-run step uploaded a per-job ledger under
# the default cascade-run-ledger-* name; reconcile globs and merges them.
cascade-ref: main
Loading