From 389df7c3d556e1069de0d51c3208d909c11008cc Mon Sep 17 00:00:00 2001
From: Albrecht Degering <albrecht@degering.name>
Date: Wed, 5 Aug 2026 21:55:32 +0200
Subject: [PATCH] Drain API pods before Kubernetes shutdown

---
 docs/SCALING_AND_MULTI_HOST_DEPLOYMENT.md     |  4 ++++
 tests/test_deployment_installer.py            | 16 ++++++++++++++++
 .../deployment/govoplan_deploy/kubernetes.py  | 19 +++++++++++++++++++
 3 files changed, 39 insertions(+)

diff --git a/docs/SCALING_AND_MULTI_HOST_DEPLOYMENT.md b/docs/SCALING_AND_MULTI_HOST_DEPLOYMENT.md
index 19f0d1a..095fcc4 100644
--- a/docs/SCALING_AND_MULTI_HOST_DEPLOYMENT.md
+++ b/docs/SCALING_AND_MULTI_HOST_DEPLOYMENT.md
@@ -270,6 +270,10 @@ record under the installation evidence directory and never retains the API key.
 
 Use `--exercise-api-pod-loss` in an approved drill window to delete one API pod,
 observe the public readiness path continuously, and record its replacement.
+Generated API workloads use a ten-second pre-stop drain so Kubernetes can remove
+the terminating endpoint from ingress and service routing before Uvicorn exits.
+Do not remove or shorten this drain without repeating the public-path pod-loss
+test against the target ingress controller and network implementation.
 This proves the bounded stateless-node-loss slice only. Session continuity,
 accepted-job redelivery, state-service failover, and coordinated restore remain
 separate target exercises whose signed evidence is governed by
diff --git a/tests/test_deployment_installer.py b/tests/test_deployment_installer.py
index 4091bc7..0593505 100644
--- a/tests/test_deployment_installer.py
+++ b/tests/test_deployment_installer.py
@@ -460,6 +460,22 @@ class DeploymentInstallerTests(unittest.TestCase):
                 "containers"
             ][0]["readinessProbe"]["httpGet"]["httpHeaders"],
         )
+        api_pod_spec = deployments["govoplan-cluster-api"]["spec"]["template"][
+            "spec"
+        ]
+        self.assertEqual(30, api_pod_spec["terminationGracePeriodSeconds"])
+        self.assertEqual(
+            ["/bin/sh", "-c", "sleep 10"],
+            api_pod_spec["containers"][0]["lifecycle"]["preStop"]["exec"][
+                "command"
+            ],
+        )
+        self.assertNotIn(
+            "lifecycle",
+            deployments["govoplan-cluster-worker"]["spec"]["template"]["spec"][
+                "containers"
+            ][0],
+        )
         worker_command = deployments["govoplan-cluster-worker"]["spec"]["template"][
             "spec"
         ]["containers"][0]["command"]
diff --git a/tools/deployment/govoplan_deploy/kubernetes.py b/tools/deployment/govoplan_deploy/kubernetes.py
index f95c473..b0e9279 100644
--- a/tools/deployment/govoplan_deploy/kubernetes.py
+++ b/tools/deployment/govoplan_deploy/kubernetes.py
@@ -168,6 +168,7 @@ def render_kubernetes(
             readiness_path="/health/ready",
             liveness_path="/health",
             probe_host=public_host,
+            graceful_shutdown_seconds=10,
             extra_environment=_role_database_environment(environment, "API"),
         ),
         _service(
@@ -658,6 +659,7 @@ def _deployment(
     readiness_path: str | None = None,
     liveness_path: str | None = None,
     probe_host: str | None = None,
+    graceful_shutdown_seconds: int = 0,
     extra_environment: Mapping[str, str] | None = None,
     selector_labels: Mapping[str, str] | None = None,
 ) -> dict[str, Any]:
@@ -715,6 +717,18 @@ def _deployment(
             container_port or 8000,
             host=probe_host,
         )
+    if graceful_shutdown_seconds:
+        container["lifecycle"] = {
+            "preStop": {
+                "exec": {
+                    "command": [
+                        "/bin/sh",
+                        "-c",
+                        f"sleep {graceful_shutdown_seconds}",
+                    ]
+                }
+            }
+        }
     pod_spec: dict[str, Any] = {
         "serviceAccountName": service_account,
         "automountServiceAccountToken": False,
@@ -734,6 +748,11 @@ def _deployment(
             }
         ],
     }
+    if graceful_shutdown_seconds:
+        pod_spec["terminationGracePeriodSeconds"] = max(
+            30,
+            graceful_shutdown_seconds + 20,
+        )
     container["volumeMounts"] = [{"name": "tmp", "mountPath": "/tmp"}]
     if s3_ca_secret_name:
         pod_spec["volumes"].append(_s3_ca_volume(s3_ca_secret_name))