chore(deploy): add migration invariant audit

deployment/device-control-core-migration-replay-audit-v1.json

@@ -0,0 +1,29 @@
+{
+  "schemaVersion": "nodedc.device-plane.device-control-core-migration-replay-audit.v1",
+  "mode": "rejected-recovery-044-live-invariants-read-only-audit",
+  "allowedOperation": "canonical-plan-only",
+  "applyAllowed": false,
+  "rejectedRecovery": {
+    "patchId": "device-control-core-migration-replay-recovery-20260822-044",
+    "artifactSha256": "b893d8c90f98943797d32f486d6477d58a3be69eb1291e28c4a4bbd2e96774b7"
+  },
+  "readOnlyEvidence": [
+    "invalid-command-kind-count",
+    "triggering-receipt-count",
+    "constraint-validated",
+    "constraint-covers-final-command-kinds",
+    "host-telemetry-table-absent"
+  ],
+  "runtimeMutation": "none",
+  "sourceMutation": "none",
+  "databaseMutation": "none",
+  "networkMutation": "none",
+  "secretRead": "none",
+  "preservedServices": [
+    "device-control-core",
+    "device-manager",
+    "device-gateway",
+    "device-postgres",
+    "device-backhaul-target"
+  ]
+}

infra/deploy-runner/build-device-control-core-migration-replay-audit-artifact.mjs

@@ -0,0 +1,95 @@
+#!/usr/bin/env node
+import { createHash } from "node:crypto";
+import { spawnSync } from "node:child_process";
+import { cp, mkdir, mkdtemp, readFile, rm, writeFile } from "node:fs/promises";
+import { tmpdir } from "node:os";
+import { dirname, join, resolve } from "node:path";
+import { fileURLToPath } from "node:url";
+
+const scriptDir = dirname(fileURLToPath(import.meta.url));
+const sourceRoot = resolve(scriptDir, "../..");
+const artifactDir = resolve(
+  process.env.NODEDC_DEPLOY_ARTIFACT_DIR
+    || resolve(scriptDir, "../deploy-artifacts"),
+);
+const [
+  patchId = "device-control-core-migration-replay-audit-20260822-045",
+  ...extra
+] = process.argv.slice(2);
+if (extra.length || !/^[A-Za-z0-9._-]{1,96}$/.test(patchId)) {
+  throw new Error(
+    "usage: build-device-control-core-migration-replay-audit-artifact.mjs [patch-id]",
+  );
+}
+
+const entry = "deployment/device-control-core-migration-replay-audit-v1.json";
+const stage = await mkdtemp(join(tmpdir(), "nodedc-control-core-migration-audit-"));
+const payload = join(stage, "payload");
+const target = resolve(artifactDir, `nodedc-device-plane-${patchId}.tgz`);
+
+try {
+  const descriptor = JSON.parse(await readFile(resolve(sourceRoot, entry), "utf8"));
+  if (
+    descriptor.schemaVersion
+      !== "nodedc.device-plane.device-control-core-migration-replay-audit.v1"
+    || descriptor.mode
+      !== "rejected-recovery-044-live-invariants-read-only-audit"
+    || descriptor.allowedOperation !== "canonical-plan-only"
+    || descriptor.applyAllowed !== false
+    || descriptor.runtimeMutation !== "none"
+    || descriptor.sourceMutation !== "none"
+    || descriptor.databaseMutation !== "none"
+    || descriptor.networkMutation !== "none"
+    || descriptor.secretRead !== "none"
+  ) {
+    throw new Error("device_control_core_migration_replay_audit_descriptor_mismatch");
+  }
+
+  await mkdir(dirname(join(payload, entry)), { recursive: true });
+  await cp(resolve(sourceRoot, entry), join(payload, entry), { force: true });
+  await writeFile(
+    join(stage, "manifest.env"),
+    `id=${patchId}\ncomponent=device-plane\ntype=app-overlay\n`,
+    "utf8",
+  );
+  await writeFile(join(stage, "files.txt"), `${entry}\n`, "utf8");
+  await mkdir(artifactDir, { recursive: true });
+  const tar = spawnSync(
+    "python3",
+    ["-c", canonicalTarScript(), target, stage],
+    { encoding: "utf8", maxBuffer: 16 * 1024 * 1024 },
+  );
+  if (tar.status !== 0) throw new Error(`tar_failed:${tar.stderr || tar.stdout}`);
+  const sha256 = createHash("sha256")
+    .update(await readFile(target))
+    .digest("hex");
+  console.log(JSON.stringify({
+    ok: true,
+    patchId,
+    component: "device-plane",
+    artifact: target,
+    sha256,
+    entries: [entry],
+    build: [],
+    services: [],
+    allowedOperation: descriptor.allowedOperation,
+    runtimeMutation: descriptor.runtimeMutation,
+  }, null, 2));
+} finally {
+  await rm(stage, { recursive: true, force: true });
+}
+
+function canonicalTarScript() {
+  return [
+    "import gzip,io,pathlib,sys,tarfile",
+    "root=pathlib.Path(sys.argv[2])",
+    "with open(sys.argv[1],'wb') as out:",
+    " with gzip.GzipFile(filename='',mode='wb',fileobj=out,compresslevel=9,mtime=0) as gz:",
+    "  with tarfile.open(fileobj=gz,mode='w',format=tarfile.PAX_FORMAT) as tar:",
+    "   for top in ('manifest.env','files.txt','payload'):",
+    "    p=root/top; paths=[p]+(sorted(p.rglob('*')) if p.is_dir() else [])",
+    "    for x in paths:",
+    "     info=tar.gettarinfo(str(x),arcname=x.relative_to(root).as_posix()); info.uid=info.gid=0; info.uname=info.gname='root'; info.mtime=0; info.mode=0o755 if info.isdir() else 0o644",
+    "     with (open(x,'rb') if info.isfile() else io.BytesIO()) as src: tar.addfile(info,src if info.isfile() else None)",
+  ].join("\n");
+}

infra/deploy-runner/build-device-control-core-release-artifact.mjs

@@ -37,6 +37,18 @@ if (
 
 const isV3 = patchId.startsWith("device-control-core-release-v3-");
 const isV2 = patchId.startsWith("device-control-core-release-v2-");
+const coreDockerfile = await readFile(
+  resolve(devicePlaneRoot, "services/device-control-core/Dockerfile"),
+  "utf8",
+);
+if (
+  !isV3
+  && coreDockerfile.includes(
+    "COPY packages/infrastructure-telemetry-contract ./packages/infrastructure-telemetry-contract",
+  )
+) {
+  throw new Error("historical_device_control_core_builder_has_advanced");
+}
 const expectedV2Predecessor = Object.freeze({
   patchId: predecessorPatchId ?? "device-control-core-release-20260812-024",
   artifactSha256: predecessorSha256 ?? "a289e909283109642e6bba3d9822a31f63423cfe0bbcd52705979681bd2bc793",

infra/deploy-runner/build-device-edge-core-channel-bootstrap-artifact.mjs

@@ -19,6 +19,17 @@ const upgradeV4 = patchId.startsWith("device-edge-core-channel-upgrade-v4-");
 const upgradeV2 = !upgradeV4 && patchId.startsWith("device-edge-core-channel-upgrade-v2-");
 const upgradeV1 = !upgradeV4 && !upgradeV2 && patchId.startsWith("device-edge-core-channel-upgrade-");
 const upgrade = upgradeV1 || upgradeV2 || upgradeV4;
+const coreDockerfile = await readFile(
+  resolve(devicePlaneRoot, "services/device-control-core/Dockerfile"),
+  "utf8",
+);
+if (
+  coreDockerfile.includes(
+    "COPY packages/infrastructure-telemetry-contract ./packages/infrastructure-telemetry-contract",
+  )
+) {
+  throw new Error("historical_device_edge_core_channel_builder_has_advanced");
+}
 const descriptorPath = upgradeV4
   ? "deployment/device-edge-core-channel-upgrade-v4.json"
   : upgradeV2

infra/deploy-runner/test_device_control_core_migration_replay_audit_artifact.py

@@ -0,0 +1,186 @@
+#!/usr/bin/env python3
+import contextlib
+import hashlib
+import importlib.machinery
+import importlib.util
+import inspect
+import io
+import json
+import os
+import re
+import subprocess
+import tarfile
+import tempfile
+import unittest
+from pathlib import Path
+from unittest import mock
+
+
+SCRIPT_DIR = Path(__file__).resolve().parent
+RUNNER_PATH = Path(
+    os.environ.get(
+        "NODEDC_DEPLOY_RUNNER",
+        SCRIPT_DIR.parents[2] / "platform/infra/deploy-runner/nodedc-deploy",
+    )
+)
+BUILDER = (
+    SCRIPT_DIR
+    / "build-device-control-core-migration-replay-audit-artifact.mjs"
+)
+
+
+def load_runner():
+    loader = importlib.machinery.SourceFileLoader(
+        "nodedc_control_core_migration_replay_audit_runner_under_test",
+        str(RUNNER_PATH),
+    )
+    spec = importlib.util.spec_from_loader(loader.name, loader)
+    module = importlib.util.module_from_spec(spec)
+    loader.exec_module(module)
+    return module
+
+
+RUNNER = load_runner()
+
+
+class DeviceControlCoreMigrationReplayAuditArtifactTest(unittest.TestCase):
+    def build(self, artifact_dir, patch_id):
+        environment = os.environ.copy()
+        environment["NODEDC_DEPLOY_ARTIFACT_DIR"] = str(artifact_dir)
+        result = subprocess.run(
+            ["node", str(BUILDER), patch_id],
+            check=True,
+            capture_output=True,
+            text=True,
+            env=environment,
+        )
+        return json.loads(result.stdout)
+
+    def test_artifact_is_deterministic_plan_only_audit(self):
+        with tempfile.TemporaryDirectory(
+            prefix="nodedc-control-core-migration-replay-audit-",
+        ) as directory:
+            artifact_dir = Path(directory)
+            patch_id = "device-control-core-migration-replay-audit-unit-045"
+            first = self.build(artifact_dir, patch_id)
+            first_bytes = Path(first["artifact"]).read_bytes()
+            second = self.build(artifact_dir, patch_id)
+            second_bytes = Path(second["artifact"]).read_bytes()
+
+            self.assertEqual(first_bytes, second_bytes)
+            self.assertEqual(
+                first["sha256"],
+                hashlib.sha256(first_bytes).hexdigest(),
+            )
+            self.assertEqual(first["build"], [])
+            self.assertEqual(first["services"], [])
+            self.assertEqual(first["allowedOperation"], "canonical-plan-only")
+            self.assertEqual(first["runtimeMutation"], "none")
+            with tarfile.open(first["artifact"], "r:gz") as archive:
+                descriptor = json.loads(
+                    archive.extractfile(
+                        "payload/"
+                        + RUNNER.DEVICE_PLANE_CONTROL_CORE_MIGRATION_REPLAY_AUDIT_REL
+                    ).read().decode("utf-8")
+                )
+            self.assertEqual(
+                descriptor,
+                RUNNER.expected_device_plane_control_core_migration_replay_audit_descriptor(),
+            )
+            entries = RUNNER.DEVICE_PLANE_CONTROL_CORE_MIGRATION_REPLAY_AUDIT_ENTRIES
+            self.assertEqual(RUNNER.component_services("device-plane", entries), ())
+            self.assertEqual(RUNNER.component_builds("device-plane", entries), ())
+
+    def test_database_audit_returns_all_mismatches_without_mutation(self):
+        database_result = mock.Mock(
+            returncode=0,
+            stdout="2\t0\tfalse\tfalse\tfalse\n",
+            stderr="",
+        )
+        with mock.patch.object(
+            RUNNER,
+            "device_plane_service_container_ids",
+            return_value=("p" * 64,),
+        ), mock.patch.object(
+            RUNNER.subprocess,
+            "run",
+            return_value=database_result,
+        ) as run:
+            evidence = (
+                RUNNER.collect_device_plane_control_core_migration_replay_database_evidence(
+                    enforce_recovery_invariants=False,
+                )
+            )
+
+        self.assertEqual(evidence["invalidCommandKindCount"], 2)
+        self.assertEqual(evidence["triggeringReceiptCount"], 0)
+        self.assertFalse(evidence["constraintValidated"])
+        self.assertFalse(evidence["constraintCoversFinalKinds"])
+        self.assertFalse(evidence["hostTelemetryTableAbsent"])
+        self.assertFalse(evidence["recovery044Ready"])
+        command = run.call_args.args[0]
+        query = command[command.index("-c") + 1]
+        self.assertTrue(query.lstrip().lower().startswith("with constraint_state"))
+        self.assertNotIn(";", query)
+        self.assertIsNone(
+            re.search(
+                r"(?im)^\s*(?:delete|update|insert|truncate|alter|drop|create)\b",
+                query,
+            )
+        )
+
+    def test_recovery_rejection_prints_five_bounded_values(self):
+        database_result = mock.Mock(
+            returncode=0,
+            stdout="0\t0\ttrue\ttrue\ttrue\n",
+            stderr="",
+        )
+        output = io.StringIO()
+        with mock.patch.object(
+            RUNNER,
+            "device_plane_service_container_ids",
+            return_value=("p" * 64,),
+        ), mock.patch.object(
+            RUNNER.subprocess,
+            "run",
+            return_value=database_result,
+        ), contextlib.redirect_stdout(output), self.assertRaises(
+            RUNNER.DeployError
+        ):
+            RUNNER.collect_device_plane_control_core_migration_replay_database_evidence()
+
+        rendered = output.getvalue()
+        for marker in (
+            "invalid_command_kind_count=0",
+            "triggering_receipt_count=0",
+            "constraint_validated=true",
+            "constraint_covers_final_kinds=true",
+            "host_telemetry_table_absent=true",
+        ):
+            self.assertIn(marker, rendered)
+
+    def test_plan_and_apply_paths_preserve_plan_only_boundary(self):
+        plan_source = inspect.getsource(RUNNER.plan_artifact)
+        apply_source = inspect.getsource(RUNNER.apply_artifact)
+        self.assertIn(
+            "device_plane_control_core_migration_replay_audit_preflight",
+            plan_source,
+        )
+        self.assertIn(
+            "emit_device_plane_control_core_migration_replay_database_evidence",
+            plan_source,
+        )
+        self.assertIn(
+            "is_device_plane_control_core_migration_replay_audit_slice",
+            apply_source,
+        )
+        self.assertLess(
+            apply_source.index(
+                "is_device_plane_control_core_migration_replay_audit_slice"
+            ),
+            apply_source.index("state_has_sha"),
+        )
+
+
+if __name__ == "__main__":
+    unittest.main()

infra/deploy-runner/test_device_manager_control_plane_artifacts.py

@@ -144,6 +144,16 @@ class DeviceManagerControlPlaneArtifactsTest(unittest.TestCase):
             == RUNNER.DEVICE_PLANE_MANAGER_RELEASE_V1_COMPOSE_SHA256
         )
 
+    def historical_control_core_builders_are_current(self):
+        dockerfile = (
+            DEVICE_CORE_ROOT / "services/device-control-core/Dockerfile"
+        ).read_text(encoding="utf-8")
+        return (
+            "COPY packages/infrastructure-telemetry-contract "
+            "./packages/infrastructure-telemetry-contract"
+            not in dockerfile
+        )
+
     def build(self, script, patch_id, artifact_dir):
         environment = os.environ.copy()
         environment["NODEDC_DEPLOY_ARTIFACT_DIR"] = str(artifact_dir)
@@ -741,7 +751,40 @@ class DeviceManagerControlPlaneArtifactsTest(unittest.TestCase):
             completed.stderr,
         )
 
+    def test_historical_core_builders_fail_closed_after_telemetry_dockerfile(self):
+        if self.historical_control_core_builders_are_current():
+            self.skipTest("historical Core Dockerfile is still current")
+        cases = (
+            (
+                "build-device-control-core-release-artifact.mjs",
+                "device-control-core-release-rebuild-forbidden-001",
+                "historical_device_control_core_builder_has_advanced",
+            ),
+            (
+                "build-device-edge-core-channel-bootstrap-artifact.mjs",
+                "device-edge-core-channel-bootstrap-rebuild-forbidden-001",
+                "historical_device_edge_core_channel_builder_has_advanced",
+            ),
+        )
+        for script, patch_id, expected_error in cases:
+            with self.subTest(script=script), tempfile.TemporaryDirectory(
+                prefix="nodedc-device-core-historical-reject-",
+            ) as directory:
+                environment = os.environ.copy()
+                environment["NODEDC_DEPLOY_ARTIFACT_DIR"] = directory
+                completed = subprocess.run(
+                    ["node", str(SCRIPT_DIR / script), patch_id],
+                    cwd=DEVICE_CORE_ROOT,
+                    env=environment,
+                    capture_output=True,
+                    text=True,
+                )
+                self.assertNotEqual(completed.returncode, 0)
+                self.assertIn(expected_error, completed.stderr)
+
     def test_edge_core_channel_bootstrap_is_core_only_and_secret_free(self):
+        if not self.historical_control_core_builders_are_current():
+            self.skipTest("historical Core channel builder generation is frozen")
         manifest, entries, names, result = self.assert_deterministic_artifact(
             "build-device-edge-core-channel-bootstrap-artifact.mjs",
             "device-edge-core-channel-bootstrap-unit-001",
@@ -778,6 +821,8 @@ class DeviceManagerControlPlaneArtifactsTest(unittest.TestCase):
         self.assertEqual(checks[0]["expected_json"]["commandTransport"], "disabled")
 
     def test_edge_core_channel_upgrade_is_core_only_and_pins_bootstrap_018(self):
+        if not self.historical_control_core_builders_are_current():
+            self.skipTest("historical Core channel builder generation is frozen")
         manifest, entries, names, result = self.assert_deterministic_artifact(
             "build-device-edge-core-channel-bootstrap-artifact.mjs",
             "device-edge-core-channel-upgrade-unit-001",
@@ -807,6 +852,8 @@ class DeviceManagerControlPlaneArtifactsTest(unittest.TestCase):
         )
 
     def test_edge_core_channel_upgrade_v2_is_core_only_and_pins_upgrade_019(self):
+        if not self.historical_control_core_builders_are_current():
+            self.skipTest("historical Core channel builder generation is frozen")
         patch_id = "device-edge-core-channel-upgrade-v2-unit-001"
         manifest, entries, names, result = self.assert_deterministic_artifact(
             "build-device-edge-core-channel-bootstrap-artifact.mjs",
@@ -853,6 +900,8 @@ class DeviceManagerControlPlaneArtifactsTest(unittest.TestCase):
         )
 
     def test_edge_core_channel_upgrade_v4_is_core_only_and_pins_upgrade_021(self):
+        if not self.historical_control_core_builders_are_current():
+            self.skipTest("historical Core channel builder generation is frozen")
         patch_id = "device-edge-core-channel-upgrade-v4-unit-001"
         manifest, entries, names, result = self.assert_deterministic_artifact(
             "build-device-edge-core-channel-bootstrap-artifact.mjs",
@@ -944,6 +993,8 @@ class DeviceManagerControlPlaneArtifactsTest(unittest.TestCase):
                     )
 
     def test_control_core_release_is_repeatable_core_only_and_compose_free(self):
+        if not self.historical_control_core_builders_are_current():
+            self.skipTest("historical Core release v1 builder generation is frozen")
         patch_id = "device-control-core-release-unit-001"
         manifest, entries, names, result = self.assert_deterministic_artifact(
             "build-device-control-core-release-artifact.mjs",
@@ -1019,6 +1070,8 @@ class DeviceManagerControlPlaneArtifactsTest(unittest.TestCase):
         )
 
     def test_control_core_release_v2_is_typed_core_only(self):
+        if not self.historical_control_core_builders_are_current():
+            self.skipTest("historical Core release v2 builder generation is frozen")
         patch_id = "device-control-core-release-v2-unit-001"
         manifest, entries, names, result = self.assert_deterministic_artifact(
             "build-device-control-core-release-artifact.mjs",
@@ -1055,6 +1108,8 @@ class DeviceManagerControlPlaneArtifactsTest(unittest.TestCase):
         self.assertEqual(descriptor["gelios"], "untouched-legacy-only")
 
     def test_control_core_release_builder_supports_release_predecessor(self):
+        if not self.historical_control_core_builders_are_current():
+            self.skipTest("historical Core release v1 builder generation is frozen")
         with tempfile.TemporaryDirectory(
             prefix="nodedc-control-core-successor-",
         ) as directory:
@@ -1111,6 +1166,8 @@ class DeviceManagerControlPlaneArtifactsTest(unittest.TestCase):
             )
 
     def test_control_core_release_v2_builder_supports_v2_release_predecessor(self):
+        if not self.historical_control_core_builders_are_current():
+            self.skipTest("historical Core release v2 builder generation is frozen")
         with tempfile.TemporaryDirectory(
             prefix="nodedc-control-core-v2-successor-",
         ) as directory:
@@ -2583,7 +2640,6 @@ class DeviceManagerControlPlaneArtifactsTest(unittest.TestCase):
                 )),
                 encoding="utf-8",
             )
-            restored_core_id = "f" * 64
             with (
                 mock.patch.object(
                     RUNNER,
@@ -2596,29 +2652,20 @@ class DeviceManagerControlPlaneArtifactsTest(unittest.TestCase):
                 ),
                 mock.patch.object(
                     RUNNER,
-                    "run_component_runtime",
+                    "retag_device_plane_control_core_image",
+                ) as retag_core_image,
+                mock.patch.object(
+                    RUNNER,
+                    "prepare_component_runtime",
+                ) as prepare_runtime,
+                mock.patch.object(
+                    RUNNER,
+                    "run_compose",
                 ) as restore_runtime,
                 mock.patch.object(
                     RUNNER,
-                    "healthcheck_compose_service_with_grace",
+                    "accept_device_plane_control_core_rollback_runtime",
-                ) as restore_health,
+                ) as rollback_acceptance,
-                mock.patch.object(
-                    RUNNER,
-                    "compose_service_container_id",
-                    return_value=restored_core_id,
-                ),
-                mock.patch.object(
-                    RUNNER,
-                    "inspect_device_plane_container",
-                    return_value={
-                        "Id": restored_core_id,
-                        "State": {
-                            "Status": "running",
-                            "Running": True,
-                            "Health": {"Status": "healthy"},
-                        },
-                    },
-                ),
                 mock.patch.object(
                     RUNNER,
                     "validate_device_manager_control_plane_runtime",
@@ -2634,20 +2681,20 @@ class DeviceManagerControlPlaneArtifactsTest(unittest.TestCase):
                 )
 
         stop.assert_not_called()
-        restore_runtime.assert_called_once_with(
+        retag_core_image.assert_called_once_with(
+            "sha256:" + "a" * 64,
+            "Device Control Core exact pre-apply rollback image",
+        )
+        prepare_runtime.assert_called_once_with(
             "device-plane",
             existing,
+        )
+        restore_runtime.assert_called_once_with(
+            "device-plane",
             ("device-control-core",),
+            existing,
         )
-        self.assertEqual(
+        rollback_acceptance.assert_called_once()
-            [call.args for call in restore_health.call_args_list],
-            [
-                ("device-plane", "device-manager"),
-                ("device-plane", "device-gateway"),
-                ("device-plane", "device-postgres"),
-                ("device-plane", "device-backhaul-target"),
-            ],
-        )
         runtime_acceptance.assert_called_once_with(
             require_edge_channel=True,
             core_network_mode="private-egress",