Mirrors/ProxMenux
1
0
Fork 0
mirror of https://github.com/MacRimi/ProxMenux.git synced 2025-12-15 00:26:23 +00:00
Code Issues Packages Projects Releases Wiki Activity

Update health_monitor.py

Browse Source
This commit is contained in:
MacRimi
2025-11-27 17:30:19 +01:00
parent 7ec8c0cea5
commit 2b8f94f457
1 changed files with 50 additions and 7 deletions
Download Patch File Download Diff File Expand all files Collapse all files

57
AppImage/scripts/health_monitor.py
View File

@@ -940,8 +940,8 @@ class HealthMonitor:
def _check_network_optimized(self) -> Dict[str, Any]: def _check_network_optimized(self) -> Dict[str, Any]:
""" """
Optimized network check - always returns status. Optimized network check - only alerts for interfaces that are actually in use.
Checks interface status and basic latency. Avoids false positives for unused physical interfaces.
""" """
try: try:
issues = [] issues = []
@@ -949,6 +949,16 @@ class HealthMonitor:
net_if_stats = psutil.net_if_stats() net_if_stats = psutil.net_if_stats()
try:
net_io_per_nic = psutil.net_io_counters(pernic=True)
except Exception:
net_io_per_nic = {}
try:
net_if_addrs = psutil.net_if_addrs()
except Exception:
net_if_addrs = {}
active_interfaces = set() active_interfaces = set()
for interface, stats in net_if_stats.items(): for interface, stats in net_if_stats.items():
@@ -957,8 +967,41 @@ class HealthMonitor:
# Check if important interface is down # Check if important interface is down
if not stats.isup: if not stats.isup:
# Consider common PVE bridge interfaces and physical NICs as important should_alert = False
if interface.startswith('vmbr') or interface.startswith('eth') or interface.startswith('ens') or interface.startswith('enp'): alert_reason = None
# Check if it's a bridge interface (always important for VMs/LXCs)
if interface.startswith('vmbr'):
should_alert = True
alert_reason = 'Bridge interface DOWN (VMs/LXCs may be affected)'
# Check if physical interface has configuration or traffic
elif interface.startswith(('eth', 'ens', 'enp', 'eno')):
# Check if interface has IP address (configured)
has_ip = False
if interface in net_if_addrs:
for addr in net_if_addrs[interface]:
if addr.family == 2: # IPv4
has_ip = True
break
# Check if interface has traffic (has been used)
has_traffic = False
if interface in net_io_per_nic:
io_stats = net_io_per_nic[interface]
# If interface has sent or received any data, it's being used
if io_stats.bytes_sent > 0 or io_stats.bytes_recv > 0:
has_traffic = True
# Only alert if interface is configured or has been used
if has_ip:
should_alert = True
alert_reason = 'Configured interface DOWN (has IP address)'
elif has_traffic:
should_alert = True
alert_reason = 'Active interface DOWN (was handling traffic)'
if should_alert:
issues.append(f'{interface} is DOWN') issues.append(f'{interface} is DOWN')
error_key = interface error_key = interface
@@ -966,18 +1009,18 @@ class HealthMonitor:
error_key=error_key, error_key=error_key,
category='network', category='network',
severity='CRITICAL', severity='CRITICAL',
reason='Interface DOWN', reason=alert_reason or 'Interface DOWN',
details={'interface': interface, 'dismissable': True} details={'interface': interface, 'dismissable': True}
) )
interface_details[interface] = { interface_details[interface] = {
'status': 'CRITICAL', 'status': 'CRITICAL',
'reason': 'Interface DOWN', 'reason': alert_reason or 'Interface DOWN',
'dismissable': True 'dismissable': True
} }
else: else:
active_interfaces.add(interface) active_interfaces.add(interface)
if interface.startswith('vmbr') or interface.startswith('eth') or interface.startswith('ens') or interface.startswith('enp'): if interface.startswith('vmbr') or interface.startswith(('eth', 'ens', 'enp', 'eno')):
health_persistence.resolve_error(interface, 'Interface recovered') health_persistence.resolve_error(interface, 'Interface recovered')
# Check connectivity (latency) # Check connectivity (latency)