Jared Vititoe
b1dd5f9cad
Adds comprehensive per-port link troubleshooting triggered from the Inspector panel when a port has an LLDP-identified server counterpart. - diagnose.py: DiagnosticsRunner with 15-section SSH command (carrier, operstate, sysfs counters, ethtool, ethtool -i/-a/-g/-S/-m, ip link, ip addr, ip route, dmesg, lldpctl); parsers for all sections; health analyzer with 14 check codes (NO_CARRIER, HALF_DUPLEX, SPEED_MISMATCH, SFP_RX_CRITICAL, CARRIER_FLAPPING, CRC_ERRORS_HIGH, LLDP_MISMATCH, etc.) - monitor.py: PulseClient now tracks last_execution_id so callers can link back to the raw Pulse execution URL - app.py: POST /api/diagnose + GET /api/diagnose/<job_id> with daemon thread background execution and 10-minute in-memory job store - inspector.html: "Run Link Diagnostics" button (shown only when LLDP host is resolvable); full results panel: health banner, physical layer, SFP/DOM with power bars, NIC error counters, collapsible ethtool -S, flow control/ring buffers, driver info, LLDP 2-col validation, collapsible dmesg, switch port summary, "View in Pulse" link - style.css: all .diag-* CSS classes with terminal aesthetic Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
361 lines
11 KiB
Python
361 lines
11 KiB
Python
"""Gandalf – Global Advanced Network Detection And Link Facilitator.
|
||
|
||
Flask web application serving the monitoring dashboard and suppression
|
||
management UI. Authentication via Authelia forward-auth headers.
|
||
All monitoring and alerting is handled by the separate monitor.py daemon.
|
||
"""
|
||
import json
|
||
import logging
|
||
import threading
|
||
import time
|
||
import uuid
|
||
from functools import wraps
|
||
|
||
from flask import Flask, jsonify, redirect, render_template, request, url_for
|
||
|
||
import db
|
||
import diagnose
|
||
from monitor import PulseClient
|
||
|
||
logging.basicConfig(
|
||
level=logging.INFO,
|
||
format='%(asctime)s %(levelname)s %(name)s %(message)s',
|
||
)
|
||
logger = logging.getLogger('gandalf.web')
|
||
|
||
app = Flask(__name__)
|
||
|
||
_cfg = None
|
||
|
||
# In-memory diagnostic job store { job_id: { status, result, created_at } }
|
||
_diag_jobs: dict = {}
|
||
_diag_lock = threading.Lock()
|
||
|
||
|
||
def _purge_old_jobs():
|
||
"""Remove jobs older than 10 minutes (called before each new job creation)."""
|
||
cutoff = time.time() - 600
|
||
with _diag_lock:
|
||
stale = [jid for jid, j in _diag_jobs.items() if j.get('created_at', 0) < cutoff]
|
||
for jid in stale:
|
||
del _diag_jobs[jid]
|
||
|
||
|
||
def _config() -> dict:
|
||
global _cfg
|
||
if _cfg is None:
|
||
with open('config.json') as f:
|
||
_cfg = json.load(f)
|
||
return _cfg
|
||
|
||
|
||
# ---------------------------------------------------------------------------
|
||
# Auth helpers
|
||
# ---------------------------------------------------------------------------
|
||
|
||
def _get_user() -> dict:
|
||
return {
|
||
'username': request.headers.get('Remote-User', ''),
|
||
'name': request.headers.get('Remote-Name', ''),
|
||
'email': request.headers.get('Remote-Email', ''),
|
||
'groups': [
|
||
g.strip()
|
||
for g in request.headers.get('Remote-Groups', '').split(',')
|
||
if g.strip()
|
||
],
|
||
}
|
||
|
||
|
||
def require_auth(f):
|
||
@wraps(f)
|
||
def wrapper(*args, **kwargs):
|
||
user = _get_user()
|
||
if not user['username']:
|
||
return (
|
||
'<h1>401 – Not authenticated</h1>'
|
||
'<p>Please access Gandalf through '
|
||
'<a href="https://auth.lotusguild.org">auth.lotusguild.org</a>.</p>',
|
||
401,
|
||
)
|
||
allowed = _config().get('auth', {}).get('allowed_groups', ['admin'])
|
||
if not any(g in allowed for g in user['groups']):
|
||
return (
|
||
f'<h1>403 – Access denied</h1>'
|
||
f'<p>Your account ({user["username"]}) is not in an allowed group '
|
||
f'({", ".join(allowed)}).</p>',
|
||
403,
|
||
)
|
||
return f(*args, **kwargs)
|
||
return wrapper
|
||
|
||
|
||
# ---------------------------------------------------------------------------
|
||
# Page routes
|
||
# ---------------------------------------------------------------------------
|
||
|
||
@app.route('/')
|
||
@require_auth
|
||
def index():
|
||
user = _get_user()
|
||
events = db.get_active_events()
|
||
summary = db.get_status_summary()
|
||
snapshot_raw = db.get_state('network_snapshot')
|
||
last_check = db.get_state('last_check', 'Never')
|
||
snapshot = json.loads(snapshot_raw) if snapshot_raw else {}
|
||
suppressions = db.get_active_suppressions()
|
||
return render_template(
|
||
'index.html',
|
||
user=user,
|
||
events=events,
|
||
summary=summary,
|
||
snapshot=snapshot,
|
||
last_check=last_check,
|
||
suppressions=suppressions,
|
||
)
|
||
|
||
|
||
@app.route('/links')
|
||
@require_auth
|
||
def links_page():
|
||
user = _get_user()
|
||
return render_template('links.html', user=user)
|
||
|
||
|
||
@app.route('/inspector')
|
||
@require_auth
|
||
def inspector():
|
||
user = _get_user()
|
||
return render_template('inspector.html', user=user)
|
||
|
||
|
||
@app.route('/suppressions')
|
||
@require_auth
|
||
def suppressions_page():
|
||
user = _get_user()
|
||
active = db.get_active_suppressions()
|
||
history = db.get_suppression_history(limit=50)
|
||
snapshot_raw = db.get_state('network_snapshot')
|
||
snapshot = json.loads(snapshot_raw) if snapshot_raw else {}
|
||
return render_template(
|
||
'suppressions.html',
|
||
user=user,
|
||
active=active,
|
||
history=history,
|
||
snapshot=snapshot,
|
||
)
|
||
|
||
|
||
# ---------------------------------------------------------------------------
|
||
# API routes
|
||
# ---------------------------------------------------------------------------
|
||
|
||
@app.route('/api/status')
|
||
@require_auth
|
||
def api_status():
|
||
return jsonify({
|
||
'summary': db.get_status_summary(),
|
||
'last_check': db.get_state('last_check', 'Never'),
|
||
'events': db.get_active_events(),
|
||
})
|
||
|
||
|
||
@app.route('/api/network')
|
||
@require_auth
|
||
def api_network():
|
||
raw = db.get_state('network_snapshot')
|
||
if raw:
|
||
try:
|
||
return jsonify(json.loads(raw))
|
||
except Exception:
|
||
pass
|
||
return jsonify({'hosts': {}, 'unifi': [], 'updated': None})
|
||
|
||
|
||
@app.route('/api/links')
|
||
@require_auth
|
||
def api_links():
|
||
raw = db.get_state('link_stats')
|
||
if raw:
|
||
try:
|
||
return jsonify(json.loads(raw))
|
||
except Exception:
|
||
pass
|
||
return jsonify({'hosts': {}, 'updated': None})
|
||
|
||
|
||
@app.route('/api/events')
|
||
@require_auth
|
||
def api_events():
|
||
return jsonify({
|
||
'active': db.get_active_events(),
|
||
'resolved': db.get_recent_resolved(hours=24, limit=30),
|
||
})
|
||
|
||
|
||
@app.route('/api/suppressions', methods=['GET'])
|
||
@require_auth
|
||
def api_get_suppressions():
|
||
return jsonify(db.get_active_suppressions())
|
||
|
||
|
||
@app.route('/api/suppressions', methods=['POST'])
|
||
@require_auth
|
||
def api_create_suppression():
|
||
user = _get_user()
|
||
data = request.get_json(silent=True) or {}
|
||
|
||
target_type = data.get('target_type', 'host')
|
||
target_name = (data.get('target_name') or '').strip()
|
||
target_detail = (data.get('target_detail') or '').strip()
|
||
reason = (data.get('reason') or '').strip()
|
||
expires_minutes = data.get('expires_minutes') # None = manual/permanent
|
||
|
||
if target_type not in ('host', 'interface', 'unifi_device', 'all'):
|
||
return jsonify({'error': 'Invalid target_type'}), 400
|
||
if target_type != 'all' and not target_name:
|
||
return jsonify({'error': 'target_name required'}), 400
|
||
if not reason:
|
||
return jsonify({'error': 'reason required'}), 400
|
||
|
||
sup_id = db.create_suppression(
|
||
target_type=target_type,
|
||
target_name=target_name,
|
||
target_detail=target_detail,
|
||
reason=reason,
|
||
suppressed_by=user['username'],
|
||
expires_minutes=int(expires_minutes) if expires_minutes else None,
|
||
)
|
||
logger.info(
|
||
f'Suppression #{sup_id} created by {user["username"]}: '
|
||
f'{target_type}/{target_name}/{target_detail} – {reason}'
|
||
)
|
||
return jsonify({'success': True, 'id': sup_id})
|
||
|
||
|
||
@app.route('/api/suppressions/<int:sup_id>', methods=['DELETE'])
|
||
@require_auth
|
||
def api_delete_suppression(sup_id: int):
|
||
user = _get_user()
|
||
db.deactivate_suppression(sup_id)
|
||
logger.info(f'Suppression #{sup_id} removed by {user["username"]}')
|
||
return jsonify({'success': True})
|
||
|
||
|
||
@app.route('/api/diagnose', methods=['POST'])
|
||
@require_auth
|
||
def api_diagnose_start():
|
||
"""Start a link diagnostic job. Returns {job_id}."""
|
||
data = request.get_json(silent=True) or {}
|
||
switch_name = (data.get('switch_name') or '').strip()
|
||
port_idx = data.get('port_idx')
|
||
|
||
if not switch_name or port_idx is None:
|
||
return jsonify({'error': 'switch_name and port_idx required'}), 400
|
||
|
||
# Look up switch + port in cached link_stats
|
||
raw = db.get_state('link_stats')
|
||
if not raw:
|
||
return jsonify({'error': 'No link_stats data available'}), 503
|
||
try:
|
||
link_data = json.loads(raw)
|
||
except Exception:
|
||
logger.error('Failed to parse link_stats JSON in /api/diagnose')
|
||
return jsonify({'error': 'Internal data error'}), 500
|
||
|
||
switches = link_data.get('unifi_switches', {})
|
||
sw = switches.get(switch_name)
|
||
if not sw:
|
||
return jsonify({'error': f'Switch "{switch_name}" not found'}), 404
|
||
|
||
# Find port by port_idx
|
||
port_data = None
|
||
for pname, pd in sw.get('ports', {}).items():
|
||
if pd.get('port_idx') == port_idx:
|
||
port_data = dict(pd)
|
||
port_data['name'] = pname
|
||
break
|
||
if not port_data:
|
||
return jsonify({'error': f'Port {port_idx} not found on switch "{switch_name}"'}), 404
|
||
|
||
# LLDP neighbor required to know which host+iface to SSH into
|
||
lldp = port_data.get('lldp')
|
||
if not lldp or not lldp.get('system_name'):
|
||
return jsonify({'error': 'No LLDP neighbor data for this port'}), 400
|
||
|
||
server_name = lldp['system_name']
|
||
lldp_port_id = lldp.get('port_id', '')
|
||
|
||
# Find matching host + interface in link_stats hosts
|
||
hosts = link_data.get('hosts', {})
|
||
server_ifaces = hosts.get(server_name)
|
||
if not server_ifaces:
|
||
return jsonify({'error': f'Host "{server_name}" not in link stats'}), 404
|
||
|
||
# Match interface by LLDP port_id (exact then fuzzy)
|
||
matched_iface = None
|
||
if lldp_port_id and lldp_port_id in server_ifaces:
|
||
matched_iface = lldp_port_id
|
||
if not matched_iface and lldp_port_id:
|
||
matched_iface = next(
|
||
(k for k in server_ifaces if lldp_port_id in k or k in lldp_port_id),
|
||
None
|
||
)
|
||
if not matched_iface:
|
||
matched_iface = next(iter(server_ifaces), None)
|
||
if not matched_iface:
|
||
return jsonify({'error': 'Cannot determine server interface'}), 400
|
||
|
||
# Resolve host IP from link_stats host data
|
||
host_ip = (server_ifaces.get(matched_iface) or {}).get('host_ip')
|
||
if not host_ip:
|
||
# Fallback: use LLDP mgmt IPs
|
||
mgmt_ips = lldp.get('mgmt_ips') or []
|
||
host_ip = mgmt_ips[0] if mgmt_ips else None
|
||
if not host_ip:
|
||
return jsonify({'error': 'Cannot determine host IP for SSH'}), 400
|
||
|
||
_purge_old_jobs()
|
||
job_id = str(uuid.uuid4())
|
||
with _diag_lock:
|
||
_diag_jobs[job_id] = {'status': 'running', 'result': None, 'created_at': time.time()}
|
||
|
||
def _run():
|
||
try:
|
||
cfg = _config()
|
||
pulse = PulseClient(cfg)
|
||
runner = diagnose.DiagnosticsRunner(pulse)
|
||
result = runner.run(host_ip, server_name, matched_iface, port_data)
|
||
except Exception as e:
|
||
logger.error(f'Diagnostic job {job_id} failed: {e}', exc_info=True)
|
||
result = {'status': 'error', 'error': str(e)}
|
||
with _diag_lock:
|
||
if job_id in _diag_jobs:
|
||
_diag_jobs[job_id]['status'] = 'done'
|
||
_diag_jobs[job_id]['result'] = result
|
||
|
||
t = threading.Thread(target=_run, daemon=True)
|
||
t.start()
|
||
|
||
return jsonify({'job_id': job_id})
|
||
|
||
|
||
@app.route('/api/diagnose/<job_id>', methods=['GET'])
|
||
@require_auth
|
||
def api_diagnose_poll(job_id: str):
|
||
"""Poll a diagnostic job. Returns {status, result}."""
|
||
with _diag_lock:
|
||
job = _diag_jobs.get(job_id)
|
||
if not job:
|
||
return jsonify({'error': 'Job not found'}), 404
|
||
return jsonify({'status': job['status'], 'result': job.get('result')})
|
||
|
||
|
||
@app.route('/health')
|
||
def health():
|
||
"""Health check endpoint (no auth)."""
|
||
return jsonify({'status': 'ok', 'service': 'gandalf'})
|
||
|
||
|
||
if __name__ == '__main__':
|
||
app.run(debug=True, host='0.0.0.0', port=5000)
|