feat(backend): reconcile running bottles' CA on gateway bring-up (0081)

Add `attach_bottled_agents_to_gateway()` as an `@abc.abstractmethod` on
`BottleBackend` and implement the first reconcile step across all three
backends: on a gateway cold boot, replace every already-running agent's
trusted CA with the freshly-minted gateway CA. Reconciling running bottles
is a backend responsibility (only the backend can enumerate its agents and
reach them — firecracker over SSH, docker/macOS over exec/cp), so making it
an abstract method keeps the fix cross-backend by construction.

The gateway rootfs is ephemeral, so a rebuild mints a new CA that every
running bottle distrusts (SSL verification fails — #510). This reconcile is
what distributes the fresh CA, so a routine gateway rebuild now doubles as a
free CA rotation. Per-bottle steps are best-effort: one unreachable or
malformed bottle is logged and skipped, never blocking the others.

Trigger — `Gateway.connect_to_orchestrator` now returns a cold-boot bool
(True when it actually (re)brought the gateway up, False when a healthy
current gateway was left untouched); each infra `ensure_running` gates the
reconcile on it, so it fires exactly on cold boot and never on an adopt.

Git-gate re-provision and egress-token restore fold into this same reconcile
on the same trigger in follow-up PRs (#516).

Refs #516. Addresses #510.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

tests/unit/test_backend_secret_reprovision.py

@@ -159,5 +159,126 @@ class TestFirecrackerReprovision(unittest.TestCase):
         restore.assert_called_once_with(client, {})
 
 
+class TestFirecrackerCaReconcile(unittest.TestCase):
+    """PRD 0081: attach_bottled_agents_to_gateway pushes the current gateway CA
+    into every running agent VM over SSH."""
+
+    def _run_dir(self, root: Path, ip: str = "10.243.0.3") -> Path:
+        run_dir = root / "demo"
+        run_dir.mkdir()
+        (run_dir / "bottle_id_ed25519").write_text("key")
+        (run_dir / "config.json").write_text(json.dumps({
+            "boot-source": {"boot_args": f"root=/dev/vda ip={ip}::gw:mask::eth0:off"}
+        }))
+        return run_dir
+
+    def test_pushes_current_ca_over_stdin_and_rebuilds_trust_store(self) -> None:
+        with tempfile.TemporaryDirectory() as tmp:
+            run_dir = self._run_dir(Path(tmp))
+            gw = Mock()
+            gw.ca_cert_pem.return_value = "PEM-DATA"
+            with patch.object(fc, "FirecrackerGateway", return_value=gw), \
+                 patch.object(fc.cleanup, "live_run_dirs", return_value=(run_dir,)), \
+                 patch.object(fc.util, "ssh_base_argv", return_value=["ssh", "guest"]), \
+                 patch.object(fc.subprocess, "run", return_value=_proc()) as run, \
+                 patch.object(fc, "info"):
+                fc.attach_bottled_agents_to_gateway()
+        # The cert goes over stdin (off argv), and the trust store is rebuilt.
+        self.assertEqual("PEM-DATA", run.call_args.kwargs["input"])
+        script = run.call_args.args[0][-1]
+        self.assertIn(fc.AGENT_CA_PATH, script)
+        self.assertIn("update-ca-certificates", script)
+
+    def test_unreachable_vm_is_skipped_not_fatal(self) -> None:
+        with tempfile.TemporaryDirectory() as tmp:
+            run_dir = self._run_dir(Path(tmp))
+            gw = Mock()
+            gw.ca_cert_pem.return_value = "PEM"
+            with patch.object(fc, "FirecrackerGateway", return_value=gw), \
+                 patch.object(fc.cleanup, "live_run_dirs", return_value=(run_dir,)), \
+                 patch.object(fc.util, "ssh_base_argv", return_value=["ssh", "guest"]), \
+                 patch.object(fc.subprocess, "run", return_value=_proc(1, stderr="down")), \
+                 patch.object(fc, "info") as info:
+                fc.attach_bottled_agents_to_gateway()  # does not raise
+        self.assertTrue(any("skipped" in c.args[0] for c in info.call_args_list))
+
+    def test_missing_gateway_ca_skips_the_whole_reconcile(self) -> None:
+        gw = Mock()
+        gw.ca_cert_pem.side_effect = fc.GatewayError("no CA yet")
+        with patch.object(fc, "FirecrackerGateway", return_value=gw), \
+             patch.object(fc.cleanup, "live_run_dirs") as live, \
+             patch.object(fc, "info"):
+            fc.attach_bottled_agents_to_gateway()
+        live.assert_not_called()  # bailed before enumerating agents
+
+
+class TestDockerCaReconcile(unittest.TestCase):
+    def test_pushes_ca_to_each_agent_container_excluding_gateway(self) -> None:
+        # network inspect lists the gateway (excluded) + one agent + a blank line.
+        inspect = _proc(stdout="bot-bottle-orch-gateway\nbot-bottle-a\n\n")
+        with patch.object(
+            docker, "run_docker",
+            side_effect=[inspect, _proc(), _proc(), _proc()],
+        ) as run, patch.object(docker.log, "info"):
+            docker.attach_bottled_agents_to_gateway(
+                "PEM", gateway_name="bot-bottle-orch-gateway")
+        argvs = [c.args[0] for c in run.call_args_list]
+        # First call enumerates; the agent then gets mkdir + cp + exec — never
+        # the gateway container.
+        self.assertNotIn(
+            "bot-bottle-orch-gateway",
+            " ".join(tok for a in argvs[1:] for tok in a),
+        )
+        cp = next(a for a in argvs if a[:2] == ["docker", "cp"])
+        self.assertEqual(f"bot-bottle-a:{docker.AGENT_CA_PATH}", cp[-1])
+        exec_argv = next(a for a in argvs if "update-ca-certificates" in a[-1])
+        self.assertIn(docker.AGENT_CA_PATH, exec_argv[-1])
+
+    def test_one_container_failure_does_not_block_others(self) -> None:
+        inspect = _proc(stdout="a\nb\n")
+        # a: mkdir fails (skip); b: mkdir/cp/exec all succeed.
+        with patch.object(
+            docker, "run_docker",
+            side_effect=[inspect, _proc(1, stderr="gone"), _proc(), _proc(), _proc()],
+        ), patch.object(docker.log, "info") as info:
+            docker.attach_bottled_agents_to_gateway("PEM", gateway_name="gw")
+        self.assertTrue(
+            any("skipped for a" in str(c.args[0]) for c in info.call_args_list)
+        )
+
+    def test_fetches_ca_from_gateway_when_not_supplied(self) -> None:
+        svc = Mock()
+        svc.gateway.return_value.ca_cert_pem.return_value = "FETCHED"
+        with patch.object(docker, "DockerInfraService", return_value=svc), \
+             patch.object(docker, "run_docker", return_value=_proc(stdout="")), \
+             patch.object(docker.log, "info"):
+            docker.attach_bottled_agents_to_gateway(network="net", gateway_name="gw")
+        svc.gateway.return_value.ca_cert_pem.assert_called_once()
+
+
+class TestMacosCaReconcile(unittest.TestCase):
+    def test_pushes_ca_to_each_running_agent_container(self) -> None:
+        agent = SimpleNamespace(slug="demo")
+        with patch.object(mac, "enumerate_active", return_value=[agent]), \
+             patch.object(mac.container_mod, "run_container_argv",
+                          return_value=_proc()) as run, \
+             patch.object(mac, "info"):
+            mac.attach_bottled_agents_to_gateway("PEM")
+        argvs = [c.args[0] for c in run.call_args_list]
+        name = f"{mac.CONTAINER_NAME_PREFIX}demo"
+        self.assertTrue(any(name in a for a in argvs))
+        self.assertTrue(
+            any("update-ca-certificates" in a[-1] for a in argvs if a[-2:-1] == ["-c"])
+        )
+        self.assertTrue(any(mac.AGENT_CA_PATH in " ".join(a) for a in argvs))
+
+    def test_enumeration_failure_is_best_effort(self) -> None:
+        with patch.object(mac, "enumerate_active",
+                          side_effect=mac.EnumerationError("failed")), \
+             patch.object(mac, "info") as info:
+            mac.attach_bottled_agents_to_gateway("PEM")  # does not raise
+        self.assertIn("skipped", info.call_args.args[0])
+
+
 if __name__ == "__main__":
     unittest.main()

tests/unit/test_docker_infra.py

@@ -29,10 +29,20 @@ class TestDockerInfraService(unittest.TestCase):
         self.orch.gateway_url.return_value = f"http://{ORCHESTRATOR_NAME}:8099"
         self.orch.mint_gateway_token.return_value = "gw.jwt"
         self.gw = MagicMock()
+        # Default to a cold boot (gateway (re)created) so the reconcile branch
+        # runs; the actual reconcile is stubbed so these composition tests stay
+        # isolated from docker.
+        self.gw.connect_to_orchestrator.return_value = True
         for name, mock in (("orchestrator", self.orch), ("gateway", self.gw)):
             p = patch.object(self.svc, name, return_value=mock)
             p.start()
             self.addCleanup(p.stop)
+        ap = patch(
+            "bot_bottle.backend.docker.backend.DockerBottleBackend"
+            ".attach_bottled_agents_to_gateway"
+        )
+        self.attach = ap.start()
+        self.addCleanup(ap.stop)
 
     def test_url_delegates_to_the_orchestrator(self) -> None:
         self.assertEqual("http://127.0.0.1:8099", self.svc.url())
@@ -58,6 +68,20 @@ class TestDockerInfraService(unittest.TestCase):
             f"http://{ORCHESTRATOR_NAME}:8099", "gw.jwt")
         self.orch.mint_gateway_token.assert_called_once()
 
+    def test_cold_boot_reconciles_running_bottles(self) -> None:
+        # A (re)created gateway (connect returns True) triggers the bring-up
+        # reconcile so running bottles re-trust the fresh gateway (PRD 0081).
+        self.gw.connect_to_orchestrator.return_value = True
+        self.svc.ensure_running()
+        self.attach.assert_called_once_with()
+
+    def test_no_reconcile_when_gateway_left_untouched(self) -> None:
+        # A healthy current gateway (connect returns False) is not a cold boot,
+        # so there is nothing to reconcile.
+        self.gw.connect_to_orchestrator.return_value = False
+        self.svc.ensure_running()
+        self.attach.assert_not_called()
+
     def test_stop_removes_both_containers(self) -> None:
         with patch(_RUN) as run:
             self.svc.stop()

tests/unit/test_firecracker_gateway.py

@@ -61,7 +61,10 @@ class TestFirecrackerGatewayConnect(unittest.TestCase):
         booted = infra_vm.InfraVm(guest_ip="10.243.255.3", private_key=Path("/k"))
         with patch.object(infra_vm, "boot_vm", return_value=booted) as boot, \
                 patch.object(infra_vm, "push_secret") as push:
-            gw.connect_to_orchestrator(_ORCH_URL, _TOKEN)
+            cold_booted = gw.connect_to_orchestrator(_ORCH_URL, _TOKEN)
+        # The VM is booted fresh here (cold-boot path only), so it always
+        # signals a cold boot → the caller reconciles running bottles (PRD 0081).
+        self.assertTrue(cold_booted)
         # Booted on the gateway link with the gateway role; the orchestrator's
         # guest IP (parsed off the URL) rides the cmdline as bb_orch.
         kw = boot.call_args.kwargs

tests/unit/test_firecracker_infra.py

@@ -40,7 +40,21 @@ class TestAccessors(unittest.TestCase):
         self.assertEqual(f"http://{ip}:{infra_vm.ORCHESTRATOR_PORT}", svc.url())
 
 
+_ATTACH = (
+    "bot_bottle.backend.firecracker.backend.FirecrackerBottleBackend"
+    ".attach_bottled_agents_to_gateway"
+)
+
+
 class TestEnsureRunning(unittest.TestCase):
+    def setUp(self) -> None:
+        # The cold-boot branch reconciles running bottles against the fresh
+        # gateway (PRD 0081). Stub it so these composition tests stay isolated
+        # from SSH; the wiring itself is asserted in test_cold_boot_reconciles*.
+        ap = patch(_ATTACH)
+        self.attach = ap.start()
+        self.addCleanup(ap.stop)
+
     def test_adopts_when_healthy_alive_and_version_matches(self):
         # Healthy control plane + live gateway + existing key + matching version
         # marker -> adopt (no stop/build/boot); returns the orchestrator URL.
@@ -59,6 +73,8 @@ class TestEnsureRunning(unittest.TestCase):
             built.assert_not_called()
             ip = infra_vm.netpool.orch_slot().guest_ip
             self.assertEqual(f"http://{ip}:{infra_vm.ORCHESTRATOR_PORT}", url)
+            # Adopt path — no cold boot, so no running-bottle reconcile.
+            self.attach.assert_not_called()
 
     def test_reboots_both_when_version_stale(self):
         # Healthy control plane but the running pair booted an OLDER image
@@ -120,6 +136,8 @@ class TestEnsureRunning(unittest.TestCase):
         # reach the control plane at startup).
         orch_cls.return_value.ensure_running.assert_called_once()
         gw_cls.return_value.connect_to_orchestrator.assert_called_once()
+        # A cold boot minted a fresh gateway CA — reconcile running bottles.
+        self.attach.assert_called_once_with()
 
 
 if __name__ == "__main__":

tests/unit/test_macos_gateway.py

@@ -44,6 +44,17 @@ class TestMacosGatewayConnect(unittest.TestCase):
     def test_default_name(self) -> None:
         self.assertEqual(GATEWAY_NAME, self.gw.name)
 
+    def test_connect_signals_cold_boot(self) -> None:
+        # The container is recreated unconditionally, so connect always signals
+        # a cold boot → the caller reconciles running bottles (PRD 0081).
+        run = Mock(return_value=_proc())
+        with patch(f"{_GW}.container_mod") as mod, \
+                patch(f"{_GW}.host_gateway_ca_dir", return_value=Path("/ca")):
+            mod.dns_server.return_value = "1.1.1.1"
+            mod.bind_mount_spec.side_effect = _spec
+            mod.run_container_argv = run
+            self.assertTrue(self.gw.connect_to_orchestrator(_ORCH_URL, _TOKEN))
+
     def test_refuses_without_orchestrator_url(self) -> None:
         with patch(f"{_GW}.container_mod") as mod:
             with self.assertRaises(GatewayError):

tests/unit/test_macos_infra.py

@@ -24,10 +24,20 @@ class TestInfraEnsureRunning(unittest.TestCase):
         self.orch.url.return_value = "http://192.168.128.2:8099"
         self.orch.mint_gateway_token.return_value = "gw.jwt"
         self.gw = MagicMock()
+        # Default to a cold boot (gateway (re)created) so the reconcile branch
+        # runs; the actual reconcile is stubbed so these composition tests stay
+        # isolated from the `container` CLI.
+        self.gw.connect_to_orchestrator.return_value = True
         for name, mock in (("orchestrator", self.orch), ("gateway", self.gw)):
             p = patch.object(self.svc, name, return_value=mock)
             p.start()
             self.addCleanup(p.stop)
+        ap = patch(
+            "bot_bottle.backend.macos_container.backend"
+            ".MacosContainerBottleBackend.attach_bottled_agents_to_gateway"
+        )
+        self.attach = ap.start()
+        self.addCleanup(ap.stop)
 
     def test_composes_networks_builds_and_brings_up_orchestrator_first(self) -> None:
         with patch(f"{_INFRA}.ensure_networks") as nets:
@@ -46,6 +56,20 @@ class TestInfraEnsureRunning(unittest.TestCase):
             "http://192.168.128.2:8099", "gw.jwt")
         self.orch.mint_gateway_token.assert_called_once()
 
+    def test_cold_boot_reconciles_running_bottles(self) -> None:
+        # A (re)created gateway (connect returns True) triggers the bring-up
+        # reconcile so running bottles re-trust the fresh gateway (PRD 0081).
+        self.gw.connect_to_orchestrator.return_value = True
+        with patch(f"{_INFRA}.ensure_networks"):
+            self.svc.ensure_running()
+        self.attach.assert_called_once_with()
+
+    def test_no_reconcile_when_gateway_left_untouched(self) -> None:
+        self.gw.connect_to_orchestrator.return_value = False
+        with patch(f"{_INFRA}.ensure_networks"):
+            self.svc.ensure_running()
+        self.attach.assert_not_called()
+
     def test_is_healthy_delegates_to_the_orchestrator(self) -> None:
         self.orch.is_healthy.return_value = True
         self.assertTrue(self.svc.is_healthy())

tests/unit/test_orchestrator_gateway.py

@@ -86,9 +86,12 @@ class TestDockerGateway(unittest.TestCase):
             return _proc()
 
         with patch(_RUN_DOCKER, side_effect=fake):
-            self.sc.connect_to_orchestrator(_ORCH_URL, _TOKEN)
+            cold_booted = self.sc.connect_to_orchestrator(_ORCH_URL, _TOKEN)
         self.assertEqual([], [c for c in calls if c[:2] == ["docker", "run"]])
         self.assertEqual([], [c for c in calls if c[:2] == ["docker", "rm"]])
+        # A healthy current gateway left untouched is not a cold boot — the
+        # bring-up flow skips the running-bottle reconcile (PRD 0081).
+        self.assertFalse(cold_booted)
 
     def test_ensure_running_recreates_when_image_is_stale(self) -> None:
         # Running, but the container was built from an OLD image → recreate so
@@ -106,9 +109,12 @@ class TestDockerGateway(unittest.TestCase):
             return _proc()
 
         with patch(_RUN_DOCKER, side_effect=fake):
-            self.sc.connect_to_orchestrator(_ORCH_URL, _TOKEN)
+            cold_booted = self.sc.connect_to_orchestrator(_ORCH_URL, _TOKEN)
         self.assertEqual(1, len([c for c in calls if c[:2] == ["docker", "run"]]))
         self.assertTrue(any(c[:2] == ["docker", "rm"] for c in calls))
+        # A recreated container minted/mounted its CA afresh — a cold boot that
+        # triggers the running-bottle reconcile (PRD 0081).
+        self.assertTrue(cold_booted)
 
     def test_ensure_running_starts_the_singleton_when_absent(self) -> None:
         calls: list[list[str]] = []
@@ -118,7 +124,8 @@ class TestDockerGateway(unittest.TestCase):
             return _proc(stdout="") if argv[:2] == ["docker", "ps"] else _proc()
 
         with patch(_RUN_DOCKER, side_effect=fake):
-            self.sc.connect_to_orchestrator(_ORCH_URL, _TOKEN)
+            cold_booted = self.sc.connect_to_orchestrator(_ORCH_URL, _TOKEN)
+        self.assertTrue(cold_booted)  # started fresh → reconcile running bottles
 
         runs = [c for c in calls if c[:2] == ["docker", "run"]]
         self.assertEqual(1, len(runs))