feat(orchestrator): update a single egress secret in place (0081)

Add a single-secret counterpart to reprovision_from_secret's restore-all:
update ONE egress token for a running bottle without a relaunch, for
refreshing a short-lived host credential (e.g. the Codex access token)
whose launch-time snapshot has expired.

- registry_store.store_agent_secret: per-key upsert (delete+insert of the
  one row), the counterpart of store_agent_secrets' replace-all.
- OrchestratorCore.update_agent_secret: set the in-memory token AND upsert
  the re-encrypted row under the bottle's env_var_secret, leaving other
  tokens untouched.
- POST /bottles/<id>/secret (cli-only) + client.update_agent_secret.

Refs #510, #512

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
This commit is contained in:
2026-07-26 18:21:40 -04:00
committed by claude
parent cecc1b4d60
commit cca724244e
8 changed files with 189 additions and 0 deletions
+21
View File
@@ -50,6 +50,12 @@ class ReprovisionBody(_StrictModel):
env_var_secret: StrictStr env_var_secret: StrictStr
class SecretBody(_StrictModel):
name: StrictStr
value: StrictStr
env_var_secret: StrictStr
class ReconcileBody(_StrictModel): class ReconcileBody(_StrictModel):
live_source_ips: list[StrictStr] live_source_ips: list[StrictStr]
grace_seconds: float | None = None grace_seconds: float | None = None
@@ -259,6 +265,21 @@ def create_app(orch: OrchestratorCore, *, signing_key: str) -> FastAPI:
return {"reprovisioned": True} return {"reprovisioned": True}
raise HTTPException(404, "no stored secrets for this bottle") raise HTTPException(404, "no stored secrets for this bottle")
@app.post("/bottles/{bottle_id}/secret")
def update_secret(bottle_id: str, body: SecretBody) -> dict[str, object]:
# Update ONE egress token for a running bottle in place — the
# single-secret form of reprovision, for refreshing a short-lived host
# credential (e.g. the Codex access token) without a relaunch. cli-only
# (not in _GATEWAY_ROUTES): a bottle must never set its own tokens.
if orch.update_agent_secret(
bottle_id,
_required(body.name, "name"),
_required(body.value, "value"),
_required(body.env_var_secret, "env_var_secret"),
):
return {"updated": True}
raise HTTPException(404, "no such bottle")
@app.delete("/bottles/{bottle_id}") @app.delete("/bottles/{bottle_id}")
def teardown(bottle_id: str) -> dict[str, object]: def teardown(bottle_id: str) -> dict[str, object]:
if orch.teardown_bottle(bottle_id): if orch.teardown_bottle(bottle_id):
+21
View File
@@ -184,6 +184,27 @@ class OrchestratorClient:
) )
return True return True
def update_agent_secret(
self, bottle_id: str, name: str, value: str, env_var_secret: str,
) -> bool:
"""Update ONE egress token for a running bottle in place
(`POST /bottles/<id>/secret`) — the single-secret form of
`reprovision_gateway`, for pushing a freshly-refreshed host credential
into a bottle without a relaunch. Returns True on success, False when the
orchestrator doesn't know the bottle (404)."""
status, _ = self._request(
"POST",
f"/bottles/{bottle_id}/secret",
{"name": name, "value": value, "env_var_secret": env_var_secret},
)
if status == 404:
return False
if not 200 <= status < 300:
raise OrchestratorClientError(
f"update_agent_secret {bottle_id}: HTTP {status}"
)
return True
def teardown_bottle(self, bottle_id: str) -> bool: def teardown_bottle(self, bottle_id: str) -> bool:
"""Tear a bottle down (`DELETE /bottles/<id>`). False if the """Tear a bottle down (`DELETE /bottles/<id>`). False if the
orchestrator didn't know it (404) — idempotent for cleanup paths.""" orchestrator didn't know it (404) — idempotent for cleanup paths."""
+24
View File
@@ -379,6 +379,30 @@ class OrchestratorCore:
self._tokens[bottle_id] = decrypted self._tokens[bottle_id] = decrypted
return True return True
def update_agent_secret(
self, bottle_id: str, name: str, value: str, env_var_secret: str,
) -> bool:
"""Update ONE egress token for a known bottle in place — the
single-secret form of ``reprovision_from_secret``.
Sets the in-memory token AND upserts the single re-encrypted row under
*env_var_secret* (the same key the rest of the rows are encrypted with, so
the whole set stays decryptable by a later ``reprovision_from_secret``),
leaving every other token untouched. Returns False if the bottle is
unknown.
Unlike ``reprovision_from_secret`` (which restores the values captured at
launch), this pushes a *caller-supplied* value — used to refresh a
short-lived host credential (e.g. the Codex access token) into a
still-running bottle without a relaunch."""
from .store.secret_store import encrypt_value
if self.registry.get(bottle_id) is None:
return False
self._tokens.setdefault(bottle_id, {})[name] = value
self.registry.store_agent_secret(
bottle_id, name, encrypt_value(env_var_secret, value))
return True
# --- consolidated gateway ---------------------------------------------- # --- consolidated gateway ----------------------------------------------
def gateway_status(self) -> dict[str, object]: def gateway_status(self) -> dict[str, object]:
@@ -370,6 +370,30 @@ class RegistryStore(DbStore):
) )
self._chmod() self._chmod()
def store_agent_secret(
self,
bottle_id: str,
key: str,
encrypted_value: str,
secret_type: str = "injected_env_var",
) -> None:
"""Upsert ONE encrypted secret row (env-var name → ciphertext) for
*bottle_id*, leaving the bottle's other secrets untouched — the per-key
counterpart of ``store_agent_secrets``' replace-all. Delete-then-insert
because the table carries no unique constraint to `ON CONFLICT` against."""
with self._connection() as conn:
conn.execute(
"DELETE FROM bottled_agent_secrets "
"WHERE bottled_agent_id = ? AND key = ? AND type = ?",
(bottle_id, key, secret_type),
)
conn.execute(
"INSERT INTO bottled_agent_secrets "
"(bottled_agent_id, key, value, type) VALUES (?, ?, ?, ?)",
(bottle_id, key, encrypted_value, secret_type),
)
self._chmod()
def get_agent_secrets( def get_agent_secrets(
self, self,
bottle_id: str, bottle_id: str,
+25
View File
@@ -151,6 +151,31 @@ class TestReprovisionGateway(unittest.TestCase):
self.c.reprovision_gateway("b1", "key") self.c.reprovision_gateway("b1", "key")
class TestUpdateAgentSecret(unittest.TestCase):
def setUp(self) -> None:
self.c = OrchestratorClient("http://orch:8080")
def test_success_posts_single_secret(self) -> None:
with patch(_URLOPEN, return_value=_resp(200, {"updated": True})) as opened:
self.assertTrue(self.c.update_agent_secret("b1", "A", "fresh", "key"))
request = opened.call_args.args[0]
self.assertEqual("POST", request.get_method())
self.assertTrue(request.full_url.endswith("/bottles/b1/secret"))
self.assertEqual(
{"name": "A", "value": "fresh", "env_var_secret": "key"},
json.loads(request.data),
)
def test_unknown_bottle_is_false(self) -> None:
with patch(_URLOPEN, side_effect=_http_error(404)):
self.assertFalse(self.c.update_agent_secret("b1", "A", "v", "key"))
def test_other_status_raises(self) -> None:
with patch(_URLOPEN, side_effect=_http_error(400)):
with self.assertRaises(OrchestratorClientError):
self.c.update_agent_secret("b1", "A", "v", "key")
class TestHealthAndPolicy(unittest.TestCase): class TestHealthAndPolicy(unittest.TestCase):
def setUp(self) -> None: def setUp(self) -> None:
self.c = OrchestratorClient("http://orch:8080") self.c = OrchestratorClient("http://orch:8080")
@@ -199,6 +199,20 @@ class TestAgentSecrets(unittest.TestCase):
got = self.store.get_agent_secrets("bottle-1") got = self.store.get_agent_secrets("bottle-1")
self.assertEqual({"K": "new", "K2": "v2"}, got) self.assertEqual({"K": "new", "K2": "v2"}, got)
def test_store_agent_secret_upserts_one_key_leaving_others(self) -> None:
self.store.store_agent_secrets("bottle-1", {"K": "v1", "K2": "v2"})
self.store.store_agent_secret("bottle-1", "K", "v1-new") # update existing
self.store.store_agent_secret("bottle-1", "K3", "v3") # insert new
self.assertEqual(
{"K": "v1-new", "K2": "v2", "K3": "v3"},
self.store.get_agent_secrets("bottle-1"),
)
def test_store_agent_secret_does_not_duplicate_rows(self) -> None:
self.store.store_agent_secret("bottle-1", "K", "v1")
self.store.store_agent_secret("bottle-1", "K", "v2")
self.assertEqual({"K": "v2"}, self.store.get_agent_secrets("bottle-1"))
def test_delete_removes_secrets(self) -> None: def test_delete_removes_secrets(self) -> None:
self.store.store_agent_secrets("bottle-1", {"K": "v"}) self.store.store_agent_secrets("bottle-1", {"K": "v"})
self.store.delete_agent_secrets("bottle-1") self.store.delete_agent_secrets("bottle-1")
+41
View File
@@ -125,6 +125,47 @@ class TestDispatch(unittest.TestCase):
{"EGRESS_TOKEN_0": "upstream-secret"}, self.orch.tokens_for(bottle_id), {"EGRESS_TOKEN_0": "upstream-secret"}, self.orch.tokens_for(bottle_id),
) )
def test_update_agent_secret_in_place(self) -> None:
key = base64.urlsafe_b64encode(b"unit-test-key").rstrip(b"=").decode()
status, payload = dispatch(
self.orch, "POST", "/bottles", _body({
"source_ip": "10.243.0.21",
"tokens": {"A": "old", "B": "keep"},
"env_var_secret": key,
}),
)
self.assertEqual(201, status)
bottle_id = payload["bottle_id"]
assert isinstance(bottle_id, str)
status, response = dispatch(
self.orch, "POST", f"/bottles/{bottle_id}/secret",
_body({"name": "A", "value": "fresh", "env_var_secret": key}),
)
self.assertEqual((200, {"updated": True}), (status, response))
self.assertEqual({"A": "fresh", "B": "keep"}, self.orch.tokens_for(bottle_id))
def test_update_agent_secret_validates_and_unknown_bottle(self) -> None:
status, _ = dispatch(self.orch, "POST", "/bottles/b1/secret", b"not-json")
self.assertEqual(400, status)
status, _ = dispatch(
self.orch, "POST", "/bottles/b1/secret", _body({"name": "A"}),
)
self.assertEqual(400, status) # value + env_var_secret missing
status, _ = dispatch(
self.orch, "POST", "/bottles/ghost/secret",
_body({"name": "A", "value": "v", "env_var_secret": "k"}),
)
self.assertEqual(404, status)
def test_update_agent_secret_is_cli_only(self) -> None:
# A data-plane (`gateway`) caller must not set a bottle's tokens.
status, _ = dispatch(
self.orch, "POST", "/bottles/b1/secret",
_body({"name": "A", "value": "v", "env_var_secret": "k"}),
role=ROLE_GATEWAY,
)
self.assertEqual(403, status)
def test_reprovision_validates_request_and_missing_rows(self) -> None: def test_reprovision_validates_request_and_missing_rows(self) -> None:
status, _ = dispatch( status, _ = dispatch(
self.orch, "POST", "/bottles/b1/reprovision_gateway", b"not-json", self.orch, "POST", "/bottles/b1/reprovision_gateway", b"not-json",
+19
View File
@@ -103,6 +103,25 @@ class TestOrchestrator(unittest.TestCase):
self.assertTrue(self.orch.reprovision_from_secret(rec.bottle_id, key)) self.assertTrue(self.orch.reprovision_from_secret(rec.bottle_id, key))
self.assertEqual({"EGRESS_TOKEN_0": "secret"}, self.orch.tokens_for(rec.bottle_id)) self.assertEqual({"EGRESS_TOKEN_0": "secret"}, self.orch.tokens_for(rec.bottle_id))
def test_update_agent_secret_refreshes_one_token_in_place(self) -> None:
key = new_env_var_secret()
rec = self.orch.launch_bottle(
"10.243.0.20", tokens={"A": "old-a", "B": "keep-b"}, env_var_secret=key,
)
self.assertTrue(self.orch.update_agent_secret(rec.bottle_id, "A", "new-a", key))
# In memory: A refreshed, B left untouched.
self.assertEqual({"A": "new-a", "B": "keep-b"}, self.orch.tokens_for(rec.bottle_id))
# At rest: the whole set stays decryptable with the SAME env_var_secret,
# so a later reprovision restores the refreshed value (not the launch one).
self.orch._tokens.clear()
self.assertTrue(self.orch.reprovision_from_secret(rec.bottle_id, key))
self.assertEqual({"A": "new-a", "B": "keep-b"}, self.orch.tokens_for(rec.bottle_id))
def test_update_agent_secret_unknown_bottle_is_false(self) -> None:
self.assertFalse(
self.orch.update_agent_secret("ghost", "A", "v", new_env_var_secret())
)
def test_reprovision_rejects_missing_rows_and_wrong_key(self) -> None: def test_reprovision_rejects_missing_rows_and_wrong_key(self) -> None:
self.assertFalse(self.orch.reprovision_from_secret("missing", new_env_var_secret())) self.assertFalse(self.orch.reprovision_from_secret("missing", new_env_var_secret()))
key = "AQEBAQEBAQEBAQEBAQEBAQEBAQEBAQEBAQEBAQEBAQE" key = "AQEBAQEBAQEBAQEBAQEBAQEBAQEBAQEBAQEBAQEBAQE"