Merge pull request 'fix: sichere Tuning-Synchronisation bei lokaler Ansible-Abweichung' (#16) from fix/ansible-tuning-sync-with-local-divergence into main

app.py

@@ -11,7 +11,7 @@ from fastapi.responses import JSONResponse, RedirectResponse
 from contextlib import asynccontextmanager
 
 log = logging.getLogger("butler")
-VERSION = "2.3.2"
+VERSION = "2.3.5"
 
 API_DIR = os.environ.get("API_KEY_DIR", "/data/api")
 VAULT_CACHE_DIR = os.environ.get("VAULT_CACHE_DIR", "/data/vault-cache")
@@ -278,6 +278,7 @@ async def root():
             "status": "GET /status - health of all backends",
             "overview": "GET /overview?details=false - deterministic homelab verdict for small models",
             "audit": "GET /audit - recent API calls",
+            "sysctl_audit": "GET /system/sysctl/{host} - read-only live and persistent network tuning",
         },
         "vault_items": len(_vault_cache),
     }
@@ -1049,6 +1050,7 @@ SPEEDTEST_REPO_FILES = (
     "streamscope/static/index.html",
     "streamscope/static/assets/app.css",
     "streamscope/static/assets/app.js",
+    "streamscope/static/assets/longterm-metrics.js",
 )
 
 
@@ -1185,6 +1187,76 @@ def _ssh(host, cmd, timeout=600):
     except _sp.TimeoutExpired:
         return 124, "", f"SSH command timed out after {timeout} seconds"
 
+
+SYSCTL_AUDIT_KEYS = (
+    "net.core.default_qdisc",
+    "net.core.rmem_default",
+    "net.core.rmem_max",
+    "net.core.wmem_default",
+    "net.core.wmem_max",
+    "net.core.netdev_max_backlog",
+    "net.core.somaxconn",
+    "net.ipv4.ip_forward",
+    "net.ipv4.tcp_congestion_control",
+    "net.ipv4.tcp_fastopen",
+    "net.ipv4.tcp_mtu_probing",
+    "net.ipv4.tcp_no_metrics_save",
+    "net.ipv4.tcp_rmem",
+    "net.ipv4.tcp_slow_start_after_idle",
+    "net.ipv4.tcp_window_scaling",
+    "net.ipv4.tcp_wmem",
+)
+
+
+def _sysctl_audit_command() -> str:
+    script = f'''import glob, json
+from pathlib import Path
+keys = {SYSCTL_AUDIT_KEYS!r}
+live, errors = {{}}, {{}}
+for key in keys:
+    try:
+        live[key] = Path("/proc/sys/" + key.replace(".", "/")).read_text().strip()
+    except OSError as exc:
+        errors[key] = str(exc)[:160]
+persistent = {{}}
+for path in ["/etc/sysctl.conf", *sorted(glob.glob("/etc/sysctl.d/*.conf"))]:
+    try:
+        with open(path, encoding="utf-8", errors="replace") as handle:
+            for raw in handle:
+                line = raw.split("#", 1)[0].strip()
+                if "=" not in line:
+                    continue
+                key, value = (part.strip() for part in line.split("=", 1))
+                if key in keys:
+                    persistent.setdefault(key, []).append({{"file": path, "value": value}})
+    except (FileNotFoundError, PermissionError):
+        pass
+print(json.dumps({{"live": live, "persistent": persistent, "errors": errors}}))
+'''
+    encoded = base64.b64encode(script.encode()).decode()
+    return f'python3 -c "import base64;exec(base64.b64decode(\'{encoded}\'))"'
+
+
+@app.get("/system/sysctl/{host}")
+async def system_sysctl_audit(host: str, _=Depends(_verify)):
+    if not re.fullmatch(r"[a-z0-9][a-z0-9-]{0,62}", host):
+        raise HTTPException(400, "Invalid host name")
+    if host == "vps":
+        target = VPS_SSH
+    else:
+        inventory = await asyncio.to_thread(_find_inventory_host, host)
+        if not inventory:
+            raise HTTPException(404, f"Host {host} not found")
+        target = f'{inventory["user"]}@{inventory["ip"]}'
+    rc, out, err = await asyncio.to_thread(_ssh, target, _sysctl_audit_command(), 30)
+    if rc != 0:
+        raise HTTPException(502, (err or out).strip()[-500:] or "sysctl audit failed")
+    try:
+        result = json.loads(out)
+    except json.JSONDecodeError as exc:
+        raise HTTPException(502, "sysctl audit returned invalid JSON") from exc
+    return {"host": host, **result}
+
 def _pve_auth():
     pv = _parse_kv("proxmox")
     return f"PVEAPIToken={pv.get('tokenid','')}={pv.get('secret','')}"
@@ -1495,10 +1567,47 @@ print("updated" if updated else "added")'''
 async def ansible_run(request: Request, _=Depends(_verify)):
     body = await request.json()
     hostname = body.get("limit", body.get("hostname", ""))
-    template_id = body.get("template_id", 10)
     if not hostname:
         return JSONResponse({"error": "limit/hostname required"}, status_code=400)
-    rc, out, err = _ssh(AUTOMATION1, f"cd /app-config/ansible && bash pfannkuchen.sh setup {hostname}", timeout=600)
+    action = body.get("action", "setup")
+    if action not in {"setup", "tune", "pvetune"}:
+        return JSONResponse({"error": "action must be setup, tune or pvetune"}, status_code=400)
+    if not re.fullmatch(r"[a-zA-Z0-9_.:-]+", hostname):
+        return JSONResponse({"error": "invalid hostname/limit"}, status_code=400)
+    if action in {"tune", "pvetune"}:
+        approved_files = (
+            "roles/sysctl/defaults/main.yml",
+            "roles/sysctl/tasks/main.yml",
+            "group_vars/vps/sysctl.yml",
+            "sysctl-proxmox.yaml",
+            "roles/sysctl_proxmox/tasks/main.yml",
+        )
+        file_sync = " && ".join(
+            f"git show origin/master:{path} > {path}" for path in approved_files
+        )
+        command = (
+            "cd /app-config/ansible && "
+            "git fetch origin master && "
+            f"{file_sync} && "
+            f"bash pfannkuchen.sh {action} {hostname}"
+        )
+    else:
+        command = (
+            "cd /app-config/ansible && "
+            "git pull --ff-only origin master && "
+            f"bash pfannkuchen.sh {action} {hostname}"
+        )
+    rc, out, err = _ssh(AUTOMATION1, command, timeout=600)
+    _audit("/ansible/run", "POST", 200 if rc == 0 else 502, f"{action} {hostname}")
+    if action != "setup":
+        return {
+            "status": "ok" if rc == 0 else "error",
+            "action": action,
+            "hostname": hostname,
+            "rc": rc,
+            "output": out[-4000:],
+            "error": err[-1000:] if rc != 0 else "",
+        }
     
     # After successful ansible run: sync Hawser token to Dockhand
     if rc == 0:

tests/test_app.py

@@ -42,7 +42,32 @@ def test_health_exposes_current_version():
     with TestClient(app.app) as client:
         response = client.get("/health")
     assert response.status_code == 200
-    assert response.json()["version"] == app.VERSION == "2.3.2"
+    assert response.json()["version"] == app.VERSION == "2.3.5"
+
+
+def test_sysctl_audit_reads_fixed_keys_from_inventory_host(monkeypatch):
+    payload = {
+        "live": {"net.ipv4.tcp_congestion_control": "bbr"},
+        "persistent": {"net.ipv4.tcp_congestion_control": [{"file": "/etc/sysctl.d/99-net-tuning.conf", "value": "bbr"}]},
+        "errors": {},
+    }
+    calls = []
+    monkeypatch.setattr(app, "_find_inventory_host", lambda name: {"name": name, "user": "root", "ip": "10.5.85.16"})
+    monkeypatch.setattr(app, "_ssh", lambda host, command, timeout=600: (calls.append((host, command, timeout)) or (0, __import__("json").dumps(payload), "")))
+    with TestClient(app.app) as client:
+        response = client.get("/system/sysctl/node6", headers={"Authorization": "Bearer test-token"})
+    assert response.status_code == 200
+    assert response.json()["live"]["net.ipv4.tcp_congestion_control"] == "bbr"
+    assert calls[0][0] == "root@10.5.85.16"
+    assert "base64.b64decode" in calls[0][1]
+
+
+def test_sysctl_audit_rejects_unknown_host_without_ssh(monkeypatch):
+    monkeypatch.setattr(app, "_find_inventory_host", lambda _name: None)
+    monkeypatch.setattr(app, "_ssh", lambda *_args, **_kwargs: (_ for _ in ()).throw(AssertionError("must not SSH")))
+    with TestClient(app.app) as client:
+        response = client.get("/system/sysctl/not-there", headers={"Authorization": "Bearer test-token"})
+    assert response.status_code == 404
 
 
 def test_invalid_log_target_is_rejected_before_ssh():
@@ -83,6 +108,46 @@ def test_inventory_upsert_uses_base64_script(monkeypatch):
     assert "\\nname =" not in calls[0][1]
 
 
+def test_ansible_run_supports_safe_tune_action_and_syncs_approved_files(monkeypatch):
+    calls = []
+
+    def fake_ssh(host, command, timeout=600):
+        calls.append((host, command, timeout))
+        return 0, "changed=1 failed=0", ""
+
+    monkeypatch.setattr(app, "_ssh", fake_ssh)
+    with TestClient(app.app) as client:
+        response = client.post(
+            "/ansible/run",
+            headers={"Authorization": "Bearer test-token"},
+            json={"hostname": "emby-sascha", "action": "tune"},
+        )
+    assert response.status_code == 200
+    assert response.json()["action"] == "tune"
+    assert "git fetch origin master" in calls[0][1]
+    assert "git show origin/master:roles/sysctl/tasks/main.yml" in calls[0][1]
+    assert "git pull --ff-only" not in calls[0][1]
+    assert "bash pfannkuchen.sh tune emby-sascha" in calls[0][1]
+    assert len(calls) == 1
+
+
+def test_ansible_run_rejects_unknown_action_and_shell_metacharacters(monkeypatch):
+    monkeypatch.setattr(app, "_ssh", lambda *_args, **_kwargs: (_ for _ in ()).throw(AssertionError("must not SSH")))
+    with TestClient(app.app) as client:
+        bad_action = client.post(
+            "/ansible/run",
+            headers={"Authorization": "Bearer test-token"},
+            json={"hostname": "emby-sascha", "action": "shell"},
+        )
+        bad_host = client.post(
+            "/ansible/run",
+            headers={"Authorization": "Bearer test-token"},
+            json={"hostname": "emby-sascha;id", "action": "tune"},
+        )
+    assert bad_action.status_code == 400
+    assert bad_host.status_code == 400
+
+
 def test_docker_inspect_returns_sanitized_summary(monkeypatch):
     raw = [{
         "Name": "/fileflows",

tests/test_streamscope_deploy.py

@@ -2,6 +2,7 @@ import app
 
 
 def test_streamscope_deploy_replaces_legacy_container_before_compose(monkeypatch):
+    assert "streamscope/static/assets/longterm-metrics.js" in app.SPEEDTEST_REPO_FILES
     files = {path: "placeholder" for path in app.SPEEDTEST_REPO_FILES}
     files["compose.yaml"] = """services:
   streamscope: