Stop long prunes failing with a bare 500 (0.60.1)

A prune that deletes a lot of data (image prune -a in particular) easily
runs past docker-py's 60s read timeout. requests' ReadTimeout is not a
DockerException, so it slipped past safe_call and the UI only showed
"Request failed with status code 500".

Prunes now use a client with a 540s timeout (below nginx's 600s
proxy_read_timeout), and safe_call turns transport timeouts and
connection errors into DockerError, so they surface as a 502 with a
message.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>

backend/docker_client.py

@@ -9,6 +9,7 @@ import functools
 from typing import Any
 
 import docker
+import requests
 from docker.errors import APIError, DockerException, NotFound
 
 from config import settings
@@ -23,19 +24,37 @@ class DockerError(Exception):
         super().__init__(f"{error}: {detail}" if detail else error)
 
 
-@functools.lru_cache(maxsize=1)
-def get_client() -> docker.DockerClient:
+# Read timeout for get_long_client(). docker-py's default of 60s is fine for
+# interactive calls, but a prune that deletes gigabytes of volume data or image
+# layers routinely takes longer. Kept below nginx's proxy_read_timeout (600s)
+# so the caller still gets our error instead of a bare gateway timeout.
+LONG_TIMEOUT_SECONDS = 540
+
+
+def _connect(**kwargs) -> docker.DockerClient:
     try:
         base_url = f"unix://{settings.DOCKER_SOCKET}"
-        client = docker.DockerClient(base_url=base_url)
+        client = docker.DockerClient(base_url=base_url, **kwargs)
         client.ping()
         return client
     except DockerException as exc:  # pragma: no cover - environment dependent
         raise DockerError("docker_unavailable", str(exc)) from exc
 
 
+@functools.lru_cache(maxsize=1)
+def get_client() -> docker.DockerClient:
+    return _connect()
+
+
+@functools.lru_cache(maxsize=1)
+def get_long_client() -> docker.DockerClient:
+    """A client for calls that can legitimately run for minutes (prunes)."""
+    return _connect(timeout=LONG_TIMEOUT_SECONDS)
+
+
 def reset_client() -> None:
     get_client.cache_clear()
+    get_long_client.cache_clear()
 
 
 def safe_call(fn, *args, **kwargs) -> Any:
@@ -48,3 +67,12 @@ def safe_call(fn, *args, **kwargs) -> Any:
         raise DockerError("docker_api_error", str(exc)) from exc
     except DockerException as exc:
         raise DockerError("docker_error", str(exc)) from exc
+    # docker-py lets transport errors from requests through unwrapped; without
+    # these a slow daemon surfaced as a bare 500 with no message.
+    except requests.exceptions.Timeout as exc:
+        raise DockerError(
+            "docker_timeout",
+            "Docker did not answer in time; the operation may still be running on the host",
+        ) from exc
+    except requests.exceptions.RequestException as exc:
+        raise DockerError("docker_unavailable", str(exc)) from exc