nexus/ProxMenux
2
0
Fork 0
mirror of https://github.com/MacRimi/ProxMenux.git synced 2026-04-18 10:02:16 +00:00
Code Issues Packages Projects Releases Wiki Activity

Update notification service

Browse Source
This commit is contained in:
MacRimi
2026-03-01 22:56:25 +01:00
parent 0dfb35730f
commit 9fe58935c4
2 changed files with 54 additions and 12 deletions
Download Patch File Download Diff File Expand all files Collapse all files

26
AppImage/scripts/health_monitor.py
View File

@@ -1500,16 +1500,22 @@ class HealthMonitor:
if sample:
reason += f'\n{sample}'
health_persistence.record_error(
error_key=error_key,
category='disks',
severity=severity,
reason=reason,
details={'disk': disk, 'device': display,
'error_count': error_count,
'smart_status': smart_health,
'sample': sample, 'dismissable': True}
)
# Only record to persistence ONCE. If the error is
# already active, don't call record_error again --
# that would keep updating last_seen and preventing
# the freshness check from detecting it as stale.
if not health_persistence.is_error_active(error_key, category='disks'):
health_persistence.record_error(
error_key=error_key,
category='disks',
severity=severity,
reason=reason,
details={'disk': disk, 'device': display,
'error_count': error_count,
'smart_status': smart_health,
'sample': sample, 'dismissable': True}
)
disk_results[display] = {
'status': severity,
'reason': reason,

40
AppImage/scripts/notification_events.py
View File

@@ -1498,18 +1498,54 @@ class PollingCollector:
self._last_notified[error_key] = now
continue
# ── Freshness check for re-notifications ──
# Don't re-notify errors whose last_seen is stale (>2h old).
# If the health monitor stopped detecting the error, last_seen
# freezes. Re-notifying with dated info is confusing.
_FRESHNESS_WINDOW = 7200 # 2 hours
last_seen_str = error.get('last_seen', '')
error_is_stale = False
if last_seen_str:
try:
from datetime import datetime as _dt
ls_epoch = _dt.fromisoformat(last_seen_str).timestamp()
if now - ls_epoch > _FRESHNESS_WINDOW:
error_is_stale = True
except (ValueError, TypeError):
pass
# Determine if we should notify
is_new = error_key not in self._known_errors
last_sent = self._last_notified.get(error_key, 0)
is_due = (now - last_sent) >= self.DIGEST_INTERVAL
if not is_new and not is_due:
continue
# For re-notifications (not new): skip if stale OR not due
if not is_new:
if error_is_stale or not is_due:
continue
# Map to our event type
event_type = self._CATEGORY_TO_EVENT_TYPE.get(category, 'system_problem')
entity, eid = self._ENTITY_MAP.get(category, ('node', ''))
# ── SMART gate for disk errors ──
# If the health monitor recorded a disk error but SMART is NOT
# FAILED, skip the notification entirely. Disk notifications
# should ONLY be sent when SMART confirms a real hardware failure.
# This prevents WARNING-level disk errors (SMART: unavailable)
# from being emitted as notifications at all.
if category in ('disks', 'smart', 'zfs'):
details = error.get('details', {})
if isinstance(details, str):
try:
details = json.loads(details)
except (json.JSONDecodeError, TypeError):
details = {}
smart_status = details.get('smart_status', '') if isinstance(details, dict) else ''
if smart_status != 'FAILED':
# SMART is PASSED, UNKNOWN, or unavailable -- don't notify
continue
# Updates are always informational notifications except
# system_age which can be WARNING (365+ days) or CRITICAL (548+ days).
emit_severity = severity