Files
bot-bottle/tests/unit/test_docker_orchestrator.py
T
didericis-codex f3fbfb3cc3
prd-number-check / require-numbered-prds (pull_request) Successful in 13s
tracker-policy-pr / check-pr (pull_request) Successful in 8s
test / integration-docker (pull_request) Failing after 35s
test / unit (pull_request) Successful in 52s
test / coverage (pull_request) Has been skipped
lint / lint (push) Successful in 1m7s
fix(ci): join control planes to the runner network
2026-07-26 08:30:57 +00:00

251 lines
11 KiB
Python

"""Unit: the docker orchestrator (control plane) container lifecycle (PRD 0070)."""
from __future__ import annotations
import unittest
import urllib.error
from pathlib import Path
from unittest.mock import MagicMock, Mock, patch
from bot_bottle.backend.docker.orchestrator import (
ORCHESTRATOR_NAME,
ORCHESTRATOR_NETWORK,
ORCHESTRATOR_SOURCE_HASH_LABEL,
DockerOrchestrator,
)
from bot_bottle.gateway import GatewayError
from bot_bottle.orchestrator.lifecycle import OrchestratorStartError, source_hash
_ORCH = "bot_bottle.backend.docker.orchestrator"
_RUN = f"{_ORCH}.run_docker"
_SLEEP = f"{_ORCH}.time.sleep"
_MONOTONIC = f"{_ORCH}.time.monotonic"
# The signing key is read through the shared provisioning contract (#476); patch
# its host-canonical key file read to keep it off the real host file.
_TOKEN = "bot_bottle.trust_domain.host_signing_key"
# The ABC's is_healthy probes /health via urllib in the lifecycle module.
_URLOPEN = "bot_bottle.orchestrator.lifecycle.urllib.request.urlopen"
def _health(status: int) -> MagicMock:
m = MagicMock()
m.__enter__.return_value.status = status
return m
def _proc(returncode: int = 0, stdout: str = "", stderr: str = "") -> Mock:
return Mock(returncode=returncode, stdout=stdout, stderr=stderr)
class TestDockerOrchestrator(unittest.TestCase):
def setUp(self) -> None:
self.orch = DockerOrchestrator(port=8099)
# _run_container seeds the signing key; keep it off the real host file.
p = patch(_TOKEN, return_value="signing-key")
p.start()
self.addCleanup(p.stop)
def test_url_is_host_loopback(self) -> None:
self.assertEqual("http://127.0.0.1:8099", self.orch.url())
def test_socket_shared_client_uses_explicit_host_and_open_bind(self) -> None:
orch = DockerOrchestrator(
port=8099, client_host="172.17.0.1", root_mount_source="state-volume"
)
with patch(_TOKEN, return_value="k"), \
patch(_URLOPEN, side_effect=[urllib.error.URLError("down"), _health(200)]), \
patch(_RUN, return_value=_proc()) as run, patch(_SLEEP):
orch.ensure_running()
self.assertEqual("http://172.17.0.1:8099", orch.url())
argv = next(c.args[0] for c in run.call_args_list
if c.args[0][:2] == ["docker", "run"])
self.assertEqual("0.0.0.0:8099:8099", argv[argv.index("--publish") + 1])
self.assertIn("state-volume:/bot-bottle-root", argv)
def test_host_path_and_named_root_mount_are_mutually_exclusive(self) -> None:
with self.assertRaisesRegex(ValueError, "host_root or root_mount_source"):
DockerOrchestrator(
host_root=Path("/host/path"),
root_mount_source="state-volume",
)
def test_socket_shared_job_network_uses_container_dns(self) -> None:
orch = DockerOrchestrator(
name="orchestrator-itest",
port=22001,
client_network="runner-job-network",
root_mount_source="state-volume",
)
with patch(_TOKEN, return_value="k"), \
patch(
_URLOPEN,
side_effect=[urllib.error.URLError("down"), _health(200)],
), patch(_RUN, return_value=_proc()) as run, patch(_SLEEP):
orch.ensure_running()
self.assertEqual("http://orchestrator-itest:8099", orch.url())
calls = [call.args[0] for call in run.call_args_list]
self.assertIn(
[
"docker", "network", "connect",
"runner-job-network", "orchestrator-itest",
],
calls,
)
argv = next(call for call in calls if call[:2] == ["docker", "run"])
self.assertEqual(
"127.0.0.1:22001:8099",
argv[argv.index("--publish") + 1],
)
def test_gateway_url_is_the_container_dns_name(self) -> None:
# The gateway reaches the orchestrator by name on the control network.
self.assertEqual(f"http://{ORCHESTRATOR_NAME}:8099", self.orch.gateway_url())
def test_is_healthy_probes_health_on_the_loopback_url(self) -> None:
with patch(_URLOPEN, return_value=_health(200)):
self.assertTrue(self.orch.is_healthy())
with patch(_URLOPEN, side_effect=urllib.error.URLError("refused")):
self.assertFalse(self.orch.is_healthy())
def test_noop_when_healthy_and_source_unchanged(self) -> None:
# A healthy orchestrator on current source is left alone — recreating it
# on every launch drops in-memory egress tokens (#381).
current = source_hash(self.orch._repo_root)
def fake(argv: list[str], **_kw: object) -> Mock:
if argv[:2] == ["docker", "ps"]:
return _proc(stdout=ORCHESTRATOR_NAME)
if argv[:2] == ["docker", "inspect"]:
return _proc(stdout=current)
return _proc()
with patch(_URLOPEN, return_value=_health(200)), \
patch(_RUN, side_effect=fake) as run, patch(_SLEEP):
self.orch.ensure_running()
self.assertEqual([], [c.args[0] for c in run.call_args_list
if c.args[0][:2] == ["docker", "run"]])
def test_recreates_when_source_changed(self) -> None:
def fake(argv: list[str], **_kw: object) -> Mock:
if argv[:2] == ["docker", "ps"]:
return _proc(stdout=ORCHESTRATOR_NAME)
if argv[:2] == ["docker", "inspect"]:
return _proc(stdout="stale-hash")
return _proc()
with patch(_URLOPEN, return_value=_health(200)), \
patch(_RUN, side_effect=fake) as run, patch(_SLEEP):
self.orch.ensure_running()
runs = [c.args[0] for c in run.call_args_list if c.args[0][:2] == ["docker", "run"]]
self.assertEqual(1, len(runs))
self.assertIn(ORCHESTRATOR_NAME, runs[0])
current = source_hash(self.orch._repo_root)
self.assertIn(f"{ORCHESTRATOR_SOURCE_HASH_LABEL}={current}", runs[0])
def test_starts_when_absent_on_control_net_and_loopback(self) -> None:
def fake(argv: list[str], **_kw: object) -> Mock:
if argv[:2] == ["docker", "ps"]:
return _proc(stdout="")
return _proc()
with patch(_URLOPEN, side_effect=[urllib.error.URLError("down"), _health(200)]), \
patch(_RUN, side_effect=fake) as run, patch(_SLEEP):
self.orch.ensure_running()
argv = next(c.args[0] for c in run.call_args_list if c.args[0][:2] == ["docker", "run"])
self.assertIn(ORCHESTRATOR_NAME, argv)
# Control network only — agents are never on it.
self.assertEqual(ORCHESTRATOR_NETWORK, argv[argv.index("--network") + 1])
# Published on loopback for the CLI, mapped to the fixed internal 8099.
self.assertEqual("127.0.0.1:8099:8099", argv[argv.index("--publish") + 1])
# The lean control plane: no mitmproxy CA mount, no gateway daemons.
self.assertFalse([a for a in argv if a.endswith(":/home/mitmproxy/.mitmproxy")])
self.assertNotIn("BOT_BOTTLE_GATEWAY_DAEMONS", " ".join(argv))
self.assertNotIn("/bot-bottle-src", " ".join(argv))
self.assertNotIn("PYTHONPATH", " ".join(argv))
# Orchestrator entrypoint args (image ENTRYPOINT is `-m bot_bottle.orchestrator`).
self.assertIn("--broker", argv)
self.assertIn("stub", argv)
def test_publish_maps_host_port_to_fixed_internal_port(self) -> None:
def fake(argv: list[str], **_kw: object) -> Mock:
return _proc(stdout="") if argv[:2] == ["docker", "ps"] else _proc()
orch = DockerOrchestrator(port=20001)
with patch(_TOKEN, return_value="k"), \
patch(_URLOPEN, side_effect=[urllib.error.URLError("down"), _health(200)]), \
patch(_RUN, side_effect=fake) as run, patch(_SLEEP):
orch.ensure_running()
argv = next(c.args[0] for c in run.call_args_list if c.args[0][:2] == ["docker", "run"])
self.assertEqual("127.0.0.1:20001:8099", argv[argv.index("--publish") + 1])
def test_raises_on_startup_timeout(self) -> None:
with patch(_URLOPEN, side_effect=urllib.error.URLError("down")), \
patch(_RUN, return_value=_proc()), \
patch(_SLEEP), patch(_MONOTONIC, side_effect=[0.0, 0.5, 2.0]):
with self.assertRaises(OrchestratorStartError):
self.orch.ensure_running(startup_timeout=1.0)
def test_noop_when_healthy_and_inspect_fails(self) -> None:
def fake(argv: list[str], **_kw: object) -> Mock:
if argv[:2] == ["docker", "ps"]:
return _proc(stdout=ORCHESTRATOR_NAME)
if argv[:2] == ["docker", "inspect"]:
return _proc(returncode=1, stderr="daemon error")
return _proc()
with patch(_URLOPEN, return_value=_health(200)), \
patch(_RUN, side_effect=fake) as run, patch(_SLEEP):
self.orch.ensure_running()
self.assertEqual([], [c.args[0] for c in run.call_args_list
if c.args[0][:2] == ["docker", "run"]])
def test_creates_control_network_internal(self) -> None:
def fake(argv: list[str], **_kw: object) -> Mock:
if argv[:3] == ["docker", "network", "inspect"]:
return _proc(returncode=1, stderr="not found")
if argv[:2] == ["docker", "ps"]:
return _proc(stdout="")
return _proc()
with patch(_URLOPEN, side_effect=[urllib.error.URLError("down"), _health(200)]), \
patch(_RUN, side_effect=fake) as run, patch(_SLEEP):
self.orch.ensure_running()
creates = [c.args[0] for c in run.call_args_list if c.args[0][:3] == ["docker", "network", "create"]]
control = [c for c in creates if ORCHESTRATOR_NETWORK in c]
self.assertEqual(1, len(control))
self.assertIn("--internal", control[0])
def test_ensure_built_builds_the_orchestrator_image(self) -> None:
with patch(_RUN, return_value=_proc()) as run:
self.orch.ensure_built()
builds = [c.args[0] for c in run.call_args_list if c.args[0][:2] == ["docker", "build"]]
self.assertEqual(1, len(builds))
self.assertTrue(any(a.endswith("Dockerfile.orchestrator") for a in builds[0]))
def test_ensure_built_raises_on_build_failure(self) -> None:
with patch(_RUN, return_value=_proc(returncode=1, stderr="no space left")):
with self.assertRaises(GatewayError):
self.orch.ensure_built()
def test_ensure_built_noop_without_dockerfile(self) -> None:
orch = DockerOrchestrator(dockerfile=None)
with patch(_RUN) as run:
orch.ensure_built()
run.assert_not_called()
def test_is_running_reads_docker_ps(self) -> None:
with patch(_RUN, return_value=_proc(stdout=ORCHESTRATOR_NAME)):
self.assertTrue(self.orch.is_running())
with patch(_RUN, return_value=_proc(stdout="")):
self.assertFalse(self.orch.is_running())
def test_stop_removes_the_container(self) -> None:
with patch(_RUN) as run:
self.orch.stop()
argv = run.call_args.args[0]
self.assertEqual(["docker", "rm", "--force", ORCHESTRATOR_NAME], argv)
if __name__ == "__main__":
unittest.main()