Compare commits

...
Author SHA1 Message Date
Ahmed Allam f143148cb5 fix(docker): drop the module docstring 2026-10-04 22:30:58 +00:00
Ahmed Allam c08614b659 fix(docker): keep DOCKER_HOST on from_env so TLS settings apply; a missing current context is an error 2026-10-04 22:19:25 +00:00
Ahmed Allam 8bc63659b7 fix(docker): use the current docker context only, no DOCKER_CONTEXT override 2026-10-04 22:08:24 +00:00
Ahmed Allam 1389ca2e33 fix(docker): drop the per-platform cause tables, show the endpoint and the raw error 2026-10-04 22:05:37 +00:00
Ahmed Allam bc5178ac67 fix(docker): connect like the docker CLI and explain why the daemon is unreachable
docker.from_env() ignores the current docker context, so Docker Desktop on
macOS (socket under ~/.docker/run unless the default-socket option is on),
OrbStack and Colima reported DOCKER NOT AVAILABLE while `docker ps` worked.
Every failure also printed the same "ensure Docker Desktop is running" text
followed by a RuntimeError traceback.

- resolve the endpoint as the CLI does: DOCKER_HOST, then DOCKER_CONTEXT or
  the current context, then the default socket; the sandbox backend uses the
  same resolution so startup and scan talk to the same daemon
- classify the SDK error (socket missing, permission denied, connection
  refused, Windows named pipe) and print the fix for the current platform,
  the endpoint that was tried and the underlying error
- exit 1 cleanly instead of raising after the panel
- telemetry reports docker_unavailable_<reason>
2026-10-04 21:52:44 +00:00
5 changed files with 311 additions and 14 deletions

No files matched your search

+1 -1
View File
@@ -150,7 +150,7 @@ When remote vars are set, Strix dual-writes telemetry to both local JSONL and th
</ParamField>
<ParamField path="DOCKER_HOST" type="string">
Docker daemon socket path. Use for remote Docker hosts or custom configurations.
Docker daemon address, for example `unix:///var/run/docker.sock` or `tcp://10.0.0.5:2375`. When not set, Strix connects to the same daemon as the `docker` CLI: the current docker context (`docker context use`), then the default socket. Docker Desktop, OrbStack and Colima register their sockets as contexts, so no extra configuration is needed for them.
</ParamField>
<ParamField path="STRIX_RUNTIME_BACKEND" default="docker" type="string">
+9 -8
View File
@@ -1599,20 +1599,21 @@ def clone_repository(repo_url: str, run_name: str, dest_name: str | None = None)
def check_docker_connection() -> Any:
import docker
from docker.errors import DockerException
from strix.runtime.docker_connection import DockerConnectionError, connect_docker
try:
return docker.from_env()
except DockerException as exc:
report_error("docker_unavailable", exc)
return connect_docker()
except DockerConnectionError as exc:
report_error("docker_unavailable", exc.cause)
console = Console()
error_text = Text()
error_text.append("DOCKER NOT AVAILABLE", style="bold red")
error_text.append("\n\n", style="white")
error_text.append("Cannot connect to Docker daemon.\n", style="white")
error_text.append(f"Cannot connect to Docker at {exc.endpoint.label}.\n", style="white")
error_text.append(f"{exc.detail}\n\n", style="dim red")
error_text.append(
"Please ensure Docker Desktop is installed and running, and try running strix again.\n",
"Make sure Docker is running. If `docker info` works in this shell, strix uses "
"the same daemon. Otherwise set DOCKER_HOST.\n",
style="white",
)
@@ -1624,7 +1625,7 @@ def check_docker_connection() -> Any:
padding=(1, 2),
)
console.print("\n", panel, "\n")
raise RuntimeError("Docker not available") from None
sys.exit(1)
def image_exists(client: Any, image_name: str) -> bool:
+6 -5
View File
@@ -27,9 +27,10 @@ async def _docker_backend(
"""Bring up a session backed by the local Docker daemon.
Uses :class:`StrixDockerSandboxClient` to inject NET_ADMIN /
NET_RAW caps + ``host.docker.internal`` host-gateway. Imports
``docker`` lazily so deployments that target a non-Docker
backend don't need the docker-py library installed.
NET_RAW caps + ``host.docker.internal`` host-gateway, on the same
endpoint the startup check resolved (``DOCKER_HOST``, docker context,
default socket). Imports ``docker`` lazily so deployments that target
a non-Docker backend don't need the docker-py library installed.
``session.start()`` is what materializes the manifest into the running
container — the SDK's ``client.create()`` only builds the inner session
@@ -37,12 +38,12 @@ async def _docker_backend(
Strix manages session lifetime explicitly via ``client.delete()`` so we
trigger ``start()`` ourselves.
"""
import docker
from agents.sandbox.sandboxes.docker import DockerSandboxClientOptions
from strix.runtime.docker_client import StrixDockerSandboxClient
from strix.runtime.docker_connection import connect_docker
client = StrixDockerSandboxClient(docker.from_env())
client = StrixDockerSandboxClient(connect_docker())
client.strix_bind_mounts = bind_mounts or []
options = DockerSandboxClientOptions(image=image, exposed_ports=exposed_ports)
session = await client.create(options=options, manifest=manifest)
+90
View File
@@ -0,0 +1,90 @@
from __future__ import annotations
import os
from dataclasses import dataclass
from typing import Any
import docker # type: ignore[import-untyped, unused-ignore]
from docker.context import ContextAPI # type: ignore[import-untyped, unused-ignore]
from docker.context.config import ( # type: ignore[import-untyped, unused-ignore]
get_current_context_name,
)
from docker.errors import DockerException # type: ignore[import-untyped, unused-ignore]
DEFAULT_CONTEXT = "default"
@dataclass(frozen=True)
class DockerEndpoint:
"""Where strix talks to Docker and how that address was chosen."""
host: str | None
source: str
tls: Any = None
@property
def label(self) -> str:
return f"{self.host} ({self.source})" if self.host else self.source
class DockerConnectionError(RuntimeError):
"""The daemon at ``endpoint`` could not be reached."""
def __init__(self, endpoint: DockerEndpoint, cause: BaseException) -> None:
self.endpoint = endpoint
self.cause = root_cause(cause)
super().__init__(f"Cannot connect to Docker at {endpoint.label}: {self.detail}")
@property
def detail(self) -> str:
text = str(self.cause).strip()
name = type(self.cause).__name__
return text if name in text else f"{name}: {text}" if text else name
def resolve_docker_endpoint(environ: dict[str, str] | None = None) -> DockerEndpoint:
"""``DOCKER_HOST``, else the current docker context, else the SDK default."""
env = os.environ if environ is None else environ
host = env.get("DOCKER_HOST", "").strip()
if host:
return DockerEndpoint(host, "DOCKER_HOST")
name = get_current_context_name()
if name == DEFAULT_CONTEXT:
return DockerEndpoint(None, "default socket")
source = f"docker context '{name}'"
try:
context = ContextAPI.get_context(name)
except DockerException as exc:
raise DockerConnectionError(DockerEndpoint(None, source), exc) from exc
if context is None:
missing = DockerException(f"{source} is selected but does not exist")
raise DockerConnectionError(DockerEndpoint(None, source), missing)
return DockerEndpoint(context.Host, source, context.TLSConfig)
def connect_docker() -> Any:
"""Return a ``docker.DockerClient`` for the resolved endpoint or raise DockerConnectionError."""
endpoint = resolve_docker_endpoint()
try:
if endpoint.source == "DOCKER_HOST" or endpoint.host is None:
return docker.from_env() # also reads DOCKER_TLS_VERIFY and DOCKER_CERT_PATH
return docker.DockerClient(base_url=endpoint.host, tls=endpoint.tls or False)
except DockerException as exc:
raise DockerConnectionError(endpoint, exc) from exc
def root_cause(exc: BaseException) -> BaseException:
"""The innermost exception: the SDK wraps the OSError in requests/urllib3 errors."""
current = exc
seen: list[BaseException] = []
while not any(current is s for s in seen):
seen.append(current)
nested = [arg for arg in current.args if isinstance(arg, BaseException)]
following = current.__cause__ or current.__context__ or (nested[-1] if nested else None)
if following is None:
break
current = following
return current
+205
View File
@@ -0,0 +1,205 @@
"""Docker endpoint resolution and the panel printed when the daemon is unreachable."""
from __future__ import annotations
import importlib
from types import SimpleNamespace
from typing import Any, cast
import pytest
from docker.errors import DockerException
from requests.exceptions import ConnectionError as RequestsConnectionError
from urllib3.exceptions import ProtocolError
from strix.runtime import backends, docker_connection
from strix.runtime.docker_connection import (
DockerConnectionError,
DockerEndpoint,
resolve_docker_endpoint,
)
cli_utils: Any = importlib.import_module("strix.interface.utils")
def _sdk_error(inner: BaseException) -> DockerException:
"""Build the exception exactly as docker-py raises it for a dead daemon."""
protocol = ProtocolError("Connection aborted.", inner)
requests_exc = RequestsConnectionError(protocol)
requests_exc.__cause__ = protocol
sdk_exc = DockerException(f"Error while fetching server API version: {requests_exc}")
sdk_exc.__cause__ = requests_exc
return sdk_exc
def test_docker_host_wins_over_context(monkeypatch: pytest.MonkeyPatch) -> None:
monkeypatch.setattr(docker_connection, "get_current_context_name", lambda: "desktop-linux")
endpoint = resolve_docker_endpoint({"DOCKER_HOST": "tcp://10.0.0.5:2375"})
assert endpoint == DockerEndpoint("tcp://10.0.0.5:2375", "DOCKER_HOST")
def test_current_context_is_used_like_the_cli(monkeypatch: pytest.MonkeyPatch) -> None:
monkeypatch.setattr(docker_connection, "get_current_context_name", lambda: "desktop-linux")
monkeypatch.setattr(
"strix.runtime.docker_connection.ContextAPI.get_context",
lambda _name: SimpleNamespace(
Host="unix:///Users/me/.docker/run/docker.sock", TLSConfig=None
),
)
endpoint = resolve_docker_endpoint({})
assert endpoint.host == "unix:///Users/me/.docker/run/docker.sock"
assert endpoint.source == "docker context 'desktop-linux'"
def test_default_context_uses_the_sdk_default(monkeypatch: pytest.MonkeyPatch) -> None:
monkeypatch.setattr(docker_connection, "get_current_context_name", lambda: "default")
assert resolve_docker_endpoint({}) == DockerEndpoint(None, "default socket")
def test_missing_or_broken_current_context_is_an_error(monkeypatch: pytest.MonkeyPatch) -> None:
monkeypatch.setattr(docker_connection, "get_current_context_name", lambda: "gone")
monkeypatch.setattr(
"strix.runtime.docker_connection.ContextAPI.get_context", lambda _name: None
)
with pytest.raises(
DockerConnectionError, match="docker context 'gone' is selected but does not exist"
):
resolve_docker_endpoint({})
def boom(_name: str) -> None:
raise DockerException("bad meta.json")
monkeypatch.setattr("strix.runtime.docker_connection.ContextAPI.get_context", boom)
with pytest.raises(DockerConnectionError, match="bad meta") as info:
resolve_docker_endpoint({})
assert info.value.endpoint.label == "docker context 'gone'"
def test_docker_host_goes_through_from_env_so_tls_settings_apply(
monkeypatch: pytest.MonkeyPatch,
) -> None:
monkeypatch.setattr(
docker_connection,
"resolve_docker_endpoint",
lambda _environ=None: DockerEndpoint("tcp://10.0.0.5:2376", "DOCKER_HOST"),
)
client = object()
monkeypatch.setattr("strix.runtime.docker_connection.docker.from_env", lambda: client)
def unexpected(**_kwargs: Any) -> None:
raise AssertionError("DOCKER_HOST must not bypass from_env")
monkeypatch.setattr("strix.runtime.docker_connection.docker.DockerClient", unexpected)
assert docker_connection.connect_docker() is client
def test_context_endpoint_keeps_its_tls_config(monkeypatch: pytest.MonkeyPatch) -> None:
tls = object()
monkeypatch.setattr(
docker_connection,
"resolve_docker_endpoint",
lambda _environ=None: DockerEndpoint("tcp://remote:2376", "docker context 'remote'", tls),
)
seen: dict[str, Any] = {}
def client(**kwargs: Any) -> str:
seen.update(kwargs)
return "client"
monkeypatch.setattr("strix.runtime.docker_connection.docker.DockerClient", client)
assert docker_connection.connect_docker() == "client"
assert seen == {"base_url": "tcp://remote:2376", "tls": tls}
@pytest.mark.parametrize(
"inner",
[
FileNotFoundError(2, "No such file or directory"),
PermissionError(13, "Permission denied"),
ConnectionRefusedError(111, "Connection refused"),
],
)
def test_connect_docker_surfaces_the_root_cause(
monkeypatch: pytest.MonkeyPatch, inner: OSError
) -> None:
monkeypatch.setattr(
docker_connection,
"resolve_docker_endpoint",
lambda _environ=None: DockerEndpoint("unix:///nope.sock", "docker context 'dead'"),
)
def dead_client(**_kwargs: Any) -> None:
raise _sdk_error(inner)
monkeypatch.setattr("strix.runtime.docker_connection.docker.DockerClient", dead_client)
with pytest.raises(DockerConnectionError) as info:
docker_connection.connect_docker()
assert info.value.cause is inner
assert info.value.detail.startswith(type(inner).__name__)
assert info.value.endpoint.host == "unix:///nope.sock"
def test_check_docker_connection_prints_endpoint_and_error_then_exits(
monkeypatch: pytest.MonkeyPatch, capsys: pytest.CaptureFixture[str]
) -> None:
reported: list[tuple[str, type | None]] = []
monkeypatch.setattr(
cli_utils, "report_error", lambda name, exc=None: reported.append((name, type(exc)))
)
error = DockerConnectionError(
DockerEndpoint(None, "default socket"), _sdk_error(PermissionError(13, "Permission denied"))
)
def failing_connect() -> None:
raise error
monkeypatch.setattr(docker_connection, "connect_docker", failing_connect)
with pytest.raises(SystemExit) as exit_info:
cli_utils.check_docker_connection()
out = capsys.readouterr().out
assert exit_info.value.code == 1
assert reported == [("docker_unavailable", PermissionError)]
assert "DOCKER NOT AVAILABLE" in out
assert "Cannot connect to Docker at default socket." in out
assert "PermissionError: [Errno 13] Permission denied" in out
assert "docker info" in out
def test_check_docker_connection_returns_the_client(monkeypatch: pytest.MonkeyPatch) -> None:
client = object()
monkeypatch.setattr(docker_connection, "connect_docker", lambda: client)
assert cli_utils.check_docker_connection() is client
@pytest.mark.asyncio
async def test_sandbox_backend_uses_the_same_endpoint(monkeypatch: pytest.MonkeyPatch) -> None:
resolved = object()
monkeypatch.setattr(docker_connection, "connect_docker", lambda: resolved)
captured: dict[str, Any] = {}
class FakeSession:
async def start(self) -> None:
captured["started"] = True
class FakeClient:
def __init__(self, docker_client: Any) -> None:
captured["docker_client"] = docker_client
async def create(self, *, options: Any, **_kwargs: Any) -> FakeSession:
captured["image"] = options.image
return FakeSession()
monkeypatch.setattr("strix.runtime.docker_client.StrixDockerSandboxClient", FakeClient)
client, session = await backends._docker_backend(
image="img:1", manifest=cast("Any", SimpleNamespace()), exposed_ports=(8080,)
)
assert captured["docker_client"] is resolved
assert captured == {"docker_client": resolved, "image": "img:1", "started": True}
assert isinstance(client, FakeClient)
assert isinstance(session, FakeSession)