test: cover lifecycle edge cases

.github/workflows/aws-runner-e2e.yml

@@ -11,6 +11,8 @@ on:
         options:
           - success
           - failure
+          - cancellation
+          - startup-timeout
           - maximum-duration
 
 permissions:
@@ -33,6 +35,7 @@ jobs:
           aws-region: ${{ vars.MICROVM_AWS_REGION }}
       - uses: ./
         id: start
+        continue-on-error: ${{ inputs.scenario == 'startup-timeout' }}
         with:
           mode: start
           github-token: ${{ secrets.RUNNER_E2E_TOKEN }}
@@ -42,10 +45,29 @@ jobs:
           cloudwatch-log-group: ${{ vars.MICROVM_RUNTIME_LOG_GROUP }}
           runner-labels: lambda-microvm,docker,e2e
           maximum-duration-seconds:
-            ${{ inputs.scenario == 'maximum-duration' && '60' || '900' }}
+            ${{ inputs.scenario == 'maximum-duration' && '60' || inputs.scenario
+            == 'cancellation' && '180' || '900' }}
+          startup-timeout-seconds:
+            ${{ inputs.scenario == 'startup-timeout' && '45' || '180' }}
+          egress-connectors:
+            ${{ inputs.scenario == 'startup-timeout' && '[]' ||
+            'INTERNET_EGRESS' }}
+      - name: Verify expected startup outcome
+        if: always()
+        env:
+          EXPECTED_FAILURE: ${{ inputs.scenario == 'startup-timeout' }}
+          START_OUTCOME: ${{ steps.start.outcome }}
+        run: |
+          if [[ "${EXPECTED_FAILURE}" == "true" ]]; then
+            test "${START_OUTCOME}" = "failure"
+          else
+            test "${START_OUTCOME}" = "success"
+          fi
 
   target:
-    if: inputs.scenario != 'maximum-duration'
+    if:
+      inputs.scenario != 'maximum-duration' && inputs.scenario !=
+      'startup-timeout'
     needs: start-runner
     runs-on: ${{ needs.start-runner.outputs.label }}
     timeout-minutes: 15
@@ -121,6 +143,10 @@ jobs:
         if: inputs.scenario == 'failure'
         run: exit 1
 
+      - name: Wait for cancellation
+        if: inputs.scenario == 'cancellation'
+        run: sleep 300
+
   duration-audit:
     if: inputs.scenario == 'maximum-duration'
     needs: start-runner

docs/testing.md

@@ -44,10 +44,12 @@ Every candidate version must prove in AWS:
 The manual `.github/workflows/aws-runner-e2e.yml` workflow exercises the
 production three-job pattern through the GitHub OIDC role. Set a temporary
 `RUNNER_E2E_TOKEN` repository secret with Administration read/write permission,
-then run its `success`, `failure`, and `maximum-duration` scenarios. Remove the
+then run its `success`, `failure`, `startup-timeout`, and `maximum-duration`
+scenarios. For `cancellation`, cancel the run while its target waits. Remove the
 secret after testing. The workflow verifies checkout, the ARM64 host, Docker,
-Buildx, Compose, service containers, DNS, egress, failure cleanup, idempotent
-stop, and the platform duration backstop.
+Buildx, Compose, service containers, DNS, egress, partial-start cleanup, job
+failure and cancellation cleanup, idempotent stop, and the platform duration
+backstop.
 
 The lower-level `.github/workflows/aws-runner-target.yml` job remains available
 for a runner launched outside GitHub Actions. It waits for a runner with the