feat: add fetch_url tool for proxying HTTP GET to cluster-internal URLs

Add a new MCP tool that lets the agent fetch URLs on the cluster's network (YARN tracking UI, Spark History Server, NodeManager web UIs) when the agent is on a different network and cannot reach those hosts directly. The MCP service runs on the YARN RM node, so it can reach every host the cluster knows about — the agent just needs a way to ask. Security: SSRF guard via host suffix overlap - URL host must share >= 2 labels of suffix with the named Connection's yarn_rm_url host (e.g. yarn_rm_url='rm.prod.internal' allows 'http://nm01.prod.internal/...') - IP literals (10.0.0.1, ::1) rejected - Non-http(s) schemes (file://, gopher://, ftp://) rejected - Connection with no yarn_rm_url cannot use this tool - Reuses Connection.auth_for_httpx() and verify_for_httpx() so the agent does not need cluster credentials - Response body capped at 1 MB (truncated=true if larger) - 30s timeout, follows redirects, loguru INFO audit log on every call - spark_executor/tools/fetch_url.py: new tool + 2 helpers (_host_suffix_overlap, _validate_url_host) - spark_executor/models.py: FetchUrlResult Pydantic model - spark_executor/tools/requests.py: FetchUrlRequest with descriptions - spark_executor/server.py: /fetch_url route, operation_id='fetch_url' - tests/unit/test_fetch_url.py: 13 unit tests covering all guards, truncation, auth/SSL pass-through, redirect follow - tests/integration/test_mcp_routes.py: assert 21 tool routes - README.md: 1 row in Spark Executor 工具 table Tests: 369 passed (up from 356). Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
2026-07-09 10:59:25 +08:00
parent f8536b63ad
commit 627e70f697
7 changed files with 334 additions and 1 deletions
@@ -64,12 +64,13 @@ def test_seventeen_tool_routes_registered():
    assert "/update_pending_job" in paths


-def test_twenty_tool_routes_registered():
+def test_twenty_one_tool_routes_registered():
    paths = {r.path for r in app.routes}
    for path in (
        "/get_external_job_logs",
        "/get_external_job_status",
        "/get_external_job_result",
+        "/fetch_url",
    ):
        assert path in paths, f"missing MCP tool route: {path}"

@@ -0,0 +1,164 @@
+# coding=utf-8
+from unittest.mock import patch
+
+import httpx
+import pytest
+
+from spark_executor.core import connection_store
+from spark_executor.core.connection_store import ConnectionStore
+from spark_executor.models import Connection
+from spark_executor.tools import connections, fetch_url
+
+
+def _fresh_stores(tmp_path, monkeypatch):
+    """Reset connection store singletons for a single test."""
+    monkeypatch.setattr(connection_store, "DEFAULT_DATA_DIR", str(tmp_path))
+    store = ConnectionStore()
+    monkeypatch.setattr(connection_store, "store", store)
+    connections.store = store
+    fetch_url.conn_store = store
+
+
+def test_fetch_url_returns_body_and_status(tmp_path, monkeypatch):
+    _fresh_stores(tmp_path, monkeypatch)
+    fetch_url.conn_store.save(
+        Connection(name="prod", master="yarn", yarn_rm_url="http://rm.prod.internal:8088")
+    )
+    resp = httpx.Response(200, text="hello", headers={"content-type": "text/html"})
+    with patch("spark_executor.tools.fetch_url.httpx.get", return_value=resp) as m:
+        out = fetch_url.fetch_url("http://nm01.prod.internal:8042/node", "prod")
+    assert out.url == "http://nm01.prod.internal:8042/node"
+    assert out.status_code == 200
+    assert out.content_type == "text/html"
+    assert out.body == "hello"
+    assert out.truncated is False
+    assert m.call_count == 1
+
+
+def test_fetch_url_raises_for_missing_connection(tmp_path, monkeypatch):
+    _fresh_stores(tmp_path, monkeypatch)
+    with pytest.raises(KeyError, match="Connection not found"):
+        fetch_url.fetch_url("http://rm.prod.internal:8088/", "missing")
+
+
+def test_fetch_url_rejects_non_http_scheme(tmp_path, monkeypatch):
+    _fresh_stores(tmp_path, monkeypatch)
+    fetch_url.conn_store.save(
+        Connection(name="prod", master="yarn", yarn_rm_url="http://rm.prod.internal:8088")
+    )
+    with pytest.raises(ValueError, match="scheme"):
+        fetch_url.fetch_url("file:///etc/passwd", "prod")
+
+
+def test_fetch_url_rejects_ftp_scheme(tmp_path, monkeypatch):
+    _fresh_stores(tmp_path, monkeypatch)
+    fetch_url.conn_store.save(
+        Connection(name="prod", master="yarn", yarn_rm_url="http://rm.prod.internal:8088")
+    )
+    with pytest.raises(ValueError, match="scheme"):
+        fetch_url.fetch_url("ftp://rm.prod.internal/foo", "prod")
+
+
+def test_fetch_url_rejects_ip_literal(tmp_path, monkeypatch):
+    _fresh_stores(tmp_path, monkeypatch)
+    fetch_url.conn_store.save(
+        Connection(name="prod", master="yarn", yarn_rm_url="http://rm.prod.internal:8088")
+    )
+    with pytest.raises(ValueError, match="IP literal"):
+        fetch_url.fetch_url("http://10.0.0.1/secrets", "prod")
+
+
+def test_fetch_url_rejects_ipv6_literal(tmp_path, monkeypatch):
+    _fresh_stores(tmp_path, monkeypatch)
+    fetch_url.conn_store.save(
+        Connection(name="prod", master="yarn", yarn_rm_url="http://rm.prod.internal:8088")
+    )
+    with pytest.raises(ValueError, match="IP literal"):
+        fetch_url.fetch_url("http://[::1]:8080/", "prod")
+
+
+def test_fetch_url_rejects_external_host(tmp_path, monkeypatch):
+    _fresh_stores(tmp_path, monkeypatch)
+    fetch_url.conn_store.save(
+        Connection(name="prod", master="yarn", yarn_rm_url="http://rm.prod.internal:8088")
+    )
+    with pytest.raises(ValueError, match="does not share enough suffix"):
+        fetch_url.fetch_url("http://evil.com/foo", "prod")
+
+
+def test_fetch_url_rejects_too_short_suffix_overlap(tmp_path, monkeypatch):
+    _fresh_stores(tmp_path, monkeypatch)
+    fetch_url.conn_store.save(
+        Connection(name="prod", master="yarn", yarn_rm_url="http://rm/8088")
+    )
+    with pytest.raises(ValueError, match="does not share enough suffix"):
+        fetch_url.fetch_url("http://other-rm/", "prod")
+
+
+def test_fetch_url_rejects_when_connection_has_no_yarn_rm_url(tmp_path, monkeypatch):
+    _fresh_stores(tmp_path, monkeypatch)
+    fetch_url.conn_store.save(Connection(name="prod", master="yarn", yarn_rm_url=None))
+    with pytest.raises(ValueError, match="no yarn_rm_url set"):
+        fetch_url.fetch_url("http://anything.com/", "prod")
+
+
+def test_fetch_url_truncates_body_over_1mb(tmp_path, monkeypatch):
+    _fresh_stores(tmp_path, monkeypatch)
+    fetch_url.conn_store.save(
+        Connection(name="prod", master="yarn", yarn_rm_url="http://rm.prod.internal:8088")
+    )
+    big = "x" * (fetch_url._MAX_BODY_BYTES + 1)
+    resp = httpx.Response(200, text=big)
+    with patch("spark_executor.tools.fetch_url.httpx.get", return_value=resp):
+        out = fetch_url.fetch_url("http://nm01.prod.internal/big", "prod")
+    assert out.truncated is True
+    assert len(out.body) == fetch_url._MAX_BODY_BYTES
+
+
+def test_fetch_url_passes_auth_from_connection(tmp_path, monkeypatch):
+    _fresh_stores(tmp_path, monkeypatch)
+    fetch_url.conn_store.save(
+        Connection(
+            name="auth",
+            master="yarn",
+            yarn_rm_url="http://rm.prod.internal:8088",
+            auth_type="basic",
+            auth_user="u",
+            auth_password="p",
+        )
+    )
+    resp = httpx.Response(200, text="ok")
+    with patch("spark_executor.tools.fetch_url.httpx.get", return_value=resp) as m:
+        fetch_url.fetch_url("http://nm01.prod.internal:8042/node", "auth")
+    auth = m.call_args.kwargs["auth"]
+    assert isinstance(auth, httpx.BasicAuth)
+    import base64
+    creds = base64.b64decode(auth._auth_header.split()[1]).decode()
+    assert creds == "u:p"
+
+
+def test_fetch_url_passes_ssl_verify_from_connection(tmp_path, monkeypatch):
+    _fresh_stores(tmp_path, monkeypatch)
+    fetch_url.conn_store.save(
+        Connection(
+            name="insecure",
+            master="yarn",
+            yarn_rm_url="http://rm.prod.internal:8088",
+            ssl_verify=False,
+        )
+    )
+    resp = httpx.Response(200, text="ok")
+    with patch("spark_executor.tools.fetch_url.httpx.get", return_value=resp) as m:
+        fetch_url.fetch_url("http://nm01.prod.internal:8042/node", "insecure")
+    assert m.call_args.kwargs["verify"] is False
+
+
+def test_fetch_url_follows_redirects(tmp_path, monkeypatch):
+    _fresh_stores(tmp_path, monkeypatch)
+    fetch_url.conn_store.save(
+        Connection(name="prod", master="yarn", yarn_rm_url="http://rm.prod.internal:8088")
+    )
+    resp = httpx.Response(200, text="ok")
+    with patch("spark_executor.tools.fetch_url.httpx.get", return_value=resp) as m:
+        fetch_url.fetch_url("http://nm01.prod.internal:8042/node", "prod")
+    assert m.call_args.kwargs["follow_redirects"] is True