Compare commits

...

4 Commits

Author SHA1 Message Date
jared 222bdb08ab Fix suppression annotation for interface_down not checking host-level rules
Lint / Python (flake8) (push) Successful in 38s
Lint / JS (eslint) (push) Successful in 7s
Security / Python Security (bandit) (push) Successful in 39s
Test / Python Tests (pytest) (push) Successful in 1m5s
Lint / Notify on failure (push) Has been skipped
Lint / Deploy (push) Successful in 4s
monitor.py checks both 'interface' and 'host' suppressions for interface_down
events, but _annotate_suppressions only checked 'interface'. A host-level
suppression would silently suppress tickets but not mark the table row as
suppressed in the UI.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
2026-05-13 13:14:46 -04:00
jared 8dd744b039 Show suppressed badge on host cards during global maintenance windows
Lint / Python (flake8) (push) Successful in 40s
Lint / JS (eslint) (push) Successful in 7s
Security / Python Security (bandit) (push) Successful in 38s
Test / Python Tests (pytest) (push) Successful in 52s
Lint / Notify on failure (push) Has been skipped
Lint / Deploy (push) Successful in 3s
Global suppressions (target_type='all') have an empty target_name, so
the selectattr filter never matched them, leaving no visual indicator
when a global maintenance window was active. Pre-compute has_global_sup
before the host loop and OR it into the badge condition.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
2026-05-13 13:12:25 -04:00
jared 9e2be150b5 Use grep -F in dmesg filter to prevent interface name treated as regex
Lint / Python (flake8) (push) Successful in 38s
Lint / JS (eslint) (push) Failing after 13s
Security / Python Security (bandit) (push) Successful in 42s
Test / Python Tests (pytest) (push) Successful in 50s
Lint / Notify on failure (push) Successful in 2s
Lint / Deploy (push) Has been skipped
grep {iface} treats dots and other special chars as regex metacharacters.
Switch to grep -F -- {iface} for fixed-string matching and to prevent
a leading dash in the interface name from being parsed as a grep flag.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
2026-05-13 11:12:02 -04:00
jared ed5ba5c59e Remove unused is_new parameter from ticket helper methods
After fixing the is_new guard bug, is_new is no longer used inside
_ticket_interface, _ticket_unifi, or _ticket_unreachable. Drop it from
their signatures and call sites.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
2026-05-13 11:10:32 -04:00
4 changed files with 28 additions and 18 deletions
+19 -10
View File
@@ -174,17 +174,26 @@ _PAGE_LIMIT = 200 # max events returned per request
def _annotate_suppressions(events: list, suppressions: list) -> None: def _annotate_suppressions(events: list, suppressions: list) -> None:
"""Annotate each event dict in-place with an is_suppressed bool.""" """Annotate each event dict in-place with an is_suppressed bool.
Mirrors the suppression check order in monitor.py exactly:
interface_down → interface OR host
unifi_device_* → unifi_device
everything else → host
"""
for ev in events: for ev in events:
sup_type = ( etype = ev.get('event_type', '')
'unifi_device' if ev.get('event_type') == 'unifi_device_offline' name = ev.get('target_name', '')
else 'interface' if ev.get('event_type') == 'interface_down' detail = ev.get('target_detail', '') or ''
else 'host' if etype == 'interface_down':
) ev['is_suppressed'] = (
ev['is_suppressed'] = db.check_suppressed( db.check_suppressed(suppressions, 'interface', name, detail) or
suppressions, sup_type, db.check_suppressed(suppressions, 'host', name)
ev.get('target_name', ''), ev.get('target_detail', '') or '', )
) elif etype == 'unifi_device_offline':
ev['is_suppressed'] = db.check_suppressed(suppressions, 'unifi_device', name, detail)
else:
ev['is_suppressed'] = db.check_suppressed(suppressions, 'host', name, detail)
# --------------------------------------------------------------------------- # ---------------------------------------------------------------------------
+1 -1
View File
@@ -68,7 +68,7 @@ class DiagnosticsRunner:
f' echo "=== ip_route ===";' f' echo "=== ip_route ===";'
f' ip route show dev {q} 2>/dev/null;' f' ip route show dev {q} 2>/dev/null;'
f' echo "=== dmesg ===";' f' echo "=== dmesg ===";'
f' dmesg 2>/dev/null | grep {q} | tail -50;' f' dmesg 2>/dev/null | grep -F -- {q} | tail -50;'
f' echo "=== lldpctl ===";' f' echo "=== lldpctl ===";'
f' lldpctl 2>/dev/null || echo "lldpd not running";' f' lldpctl 2>/dev/null || echo "lldpd not running";'
f' echo "=== end ==="' f' echo "=== end ==="'
+6 -6
View File
@@ -734,7 +734,7 @@ class NetworkMonitor:
f'Interface {iface} on {host} went link-down ({_now_utc()})', f'Interface {iface} on {host} went link-down ({_now_utc()})',
) )
if not sup and consec >= self.fail_thresh: if not sup and consec >= self.fail_thresh:
self._ticket_interface(event_id, is_new, host, iface, consec) self._ticket_interface(event_id, host, iface, consec)
if host_has_regression: if host_has_regression:
hosts_with_regression.append(host) hosts_with_regression.append(host)
@@ -771,7 +771,7 @@ class NetworkMonitor:
db.resolve_event('cluster_network_issue', self.cluster_name, '') db.resolve_event('cluster_network_issue', self.cluster_name, '')
def _ticket_interface( def _ticket_interface(
self, event_id: int, is_new: bool, host: str, iface: str, consec: int self, event_id: int, host: str, iface: str, consec: int
) -> None: ) -> None:
title = ( title = (
f'[{host}][auto][production][issue][network][single-node] ' f'[{host}][auto][production][issue][network][single-node] '
@@ -810,11 +810,11 @@ class NetworkMonitor:
f'UniFi {name} ({d.get("ip","")}) offline ({_now_utc()})', f'UniFi {name} ({d.get("ip","")}) offline ({_now_utc()})',
) )
if not sup and consec >= self.fail_thresh: if not sup and consec >= self.fail_thresh:
self._ticket_unifi(event_id, is_new, d) self._ticket_unifi(event_id, d)
else: else:
db.resolve_event('unifi_device_offline', name, d.get('type', '')) db.resolve_event('unifi_device_offline', name, d.get('type', ''))
def _ticket_unifi(self, event_id: int, is_new: bool, device: dict) -> None: def _ticket_unifi(self, event_id: int, device: dict) -> None:
name = device['name'] name = device['name']
title = ( title = (
f'[{name}][auto][production][issue][network][single-node] ' f'[{name}][auto][production][issue][network][single-node] '
@@ -850,12 +850,12 @@ class NetworkMonitor:
f'Host {name} ({ip}) unreachable via ping ({_now_utc()})', f'Host {name} ({ip}) unreachable via ping ({_now_utc()})',
) )
if not sup and consec >= self.fail_thresh: if not sup and consec >= self.fail_thresh:
self._ticket_unreachable(event_id, is_new, name, ip, consec) self._ticket_unreachable(event_id, name, ip, consec)
else: else:
db.resolve_event('host_unreachable', name, ip) db.resolve_event('host_unreachable', name, ip)
def _ticket_unreachable( def _ticket_unreachable(
self, event_id: int, is_new: bool, name: str, ip: str, consec: int self, event_id: int, name: str, ip: str, consec: int
) -> None: ) -> None:
title = ( title = (
f'[{name}][auto][production][issue][network][single-node] ' f'[{name}][auto][production][issue][network][single-node] '
+2 -1
View File
@@ -324,6 +324,7 @@
</div> </div>
</div> </div>
<div class="host-grid" id="host-grid"> <div class="host-grid" id="host-grid">
{%- set has_global_sup = suppressions | selectattr('target_type', 'equalto', 'all') | list | length > 0 -%}
{% for name, host in snapshot.hosts.items() %} {% for name, host in snapshot.hosts.items() %}
{% set suppressed = suppressions | selectattr('target_name', 'equalto', name) | list %} {% set suppressed = suppressions | selectattr('target_name', 'equalto', name) | list %}
<div class="host-card host-card-{{ host.status }}" data-host="{{ name }}"> <div class="host-card host-card-{{ host.status }}" data-host="{{ name }}">
@@ -331,7 +332,7 @@
<div class="host-name-row"> <div class="host-name-row">
<span class="host-status-dot dot-{{ host.status }}"></span> <span class="host-status-dot dot-{{ host.status }}"></span>
<span class="host-name">{{ name }}</span> <span class="host-name">{{ name }}</span>
{% if suppressed %} {% if suppressed or has_global_sup %}
<span class="badge-suppressed" title="Suppressed">🔕</span> <span class="badge-suppressed" title="Suppressed">🔕</span>
{% endif %} {% endif %}
</div> </div>