mirror of
https://github.com/MacRimi/ProxMenux.git
synced 2026-11-02 10:52:22 +00:00
fix(monitor): a permanent dismissal stays permanent, and every host backup run is notified
This commit is contained in:
@@ -0,0 +1,44 @@
|
||||
"""Every run of a host backup is reported, the way a VM or CT backup is: a
|
||||
nightly job is not the same event as the night before, and its result is not
|
||||
kept for the daily digest."""
|
||||
import sys
|
||||
from pathlib import Path
|
||||
import unittest
|
||||
|
||||
sys.path.insert(0, str(Path(__file__).resolve().parents[1]))
|
||||
import notification_manager # noqa: E402
|
||||
from notification_events import NotificationEvent # noqa: E402
|
||||
|
||||
BACKUP_EVENTS = ('backup_start', 'backup_complete', 'backup_fail',
|
||||
'host_backup_start', 'host_backup_complete', 'host_backup_fail')
|
||||
|
||||
|
||||
class HostBackupEveryRunTests(unittest.TestCase):
|
||||
def setUp(self):
|
||||
self.manager = notification_manager.NotificationManager.__new__(notification_manager.NotificationManager)
|
||||
self.manager._config = {'email.digest_enabled': 'true'}
|
||||
|
||||
def event(self, event_type, severity='INFO'):
|
||||
return NotificationEvent(event_type, severity, {'hostname': 'pve'}, source='test', entity='node', entity_id='')
|
||||
|
||||
def test_two_runs_a_day_apart_or_closer_are_both_sent(self):
|
||||
for event_type in BACKUP_EVENTS:
|
||||
with self.subTest(event_type=event_type):
|
||||
severity = 'CRITICAL' if event_type.endswith('_fail') else 'INFO'
|
||||
self.assertEqual(self.manager._cooldown_seconds(self.event(event_type, severity)), 10)
|
||||
|
||||
def test_the_result_is_not_kept_for_the_daily_digest(self):
|
||||
for event_type in BACKUP_EVENTS:
|
||||
with self.subTest(event_type=event_type):
|
||||
self.assertFalse(self.manager._should_buffer_for_digest('email', 'INFO', event_type))
|
||||
|
||||
def test_a_cooldown_set_by_the_user_is_still_respected(self):
|
||||
self.manager._config['cooldown.host_backup_complete'] = '3600'
|
||||
self.assertEqual(self.manager._cooldown_seconds(self.event('host_backup_complete')), 3600)
|
||||
|
||||
def test_other_informational_events_still_go_to_the_digest(self):
|
||||
self.assertTrue(self.manager._should_buffer_for_digest('email', 'INFO', 'update_summary'))
|
||||
|
||||
|
||||
if __name__ == '__main__':
|
||||
unittest.main()
|
||||
@@ -0,0 +1,108 @@
|
||||
"""An alert dismissed permanently stays dismissed until the user re-enables
|
||||
it: not seeing it for a while, the condition recovering or a change of the
|
||||
suppression settings does not bring it back."""
|
||||
import sqlite3
|
||||
import sys
|
||||
import tempfile
|
||||
import unittest
|
||||
from datetime import datetime, timedelta
|
||||
from pathlib import Path
|
||||
from unittest.mock import patch
|
||||
|
||||
SCRIPTS = Path(__file__).resolve().parents[1]
|
||||
sys.path.insert(0, str(SCRIPTS))
|
||||
# The module opens the database of the host when it is imported.
|
||||
_connect = sqlite3.connect
|
||||
with patch.object(Path, 'mkdir'), patch('builtins.print'), \
|
||||
patch('sqlite3.connect', side_effect=lambda *args, **kwargs: _connect(':memory:')):
|
||||
import health_persistence as hp_module
|
||||
|
||||
KEY = 'pve_storage_full_local-lvm'
|
||||
|
||||
|
||||
class PermanentDismissalTests(unittest.TestCase):
|
||||
def setUp(self):
|
||||
tmp = tempfile.TemporaryDirectory()
|
||||
self.addCleanup(tmp.cleanup)
|
||||
store = hp_module.HealthPersistence.__new__(hp_module.HealthPersistence)
|
||||
for name, value in vars(hp_module.health_persistence).items():
|
||||
setattr(store, name, value)
|
||||
store.data_dir = Path(tmp.name)
|
||||
store.db_path = store.data_dir / 'health_monitor.db'
|
||||
with patch('builtins.print'):
|
||||
store._init_database()
|
||||
self.store = store
|
||||
|
||||
def seen(self):
|
||||
return self.store.record_error(KEY, 'storage', 'WARNING', 'PVE storage local-lvm (lvmthin) at 96.0% used',
|
||||
{'storage_name': 'local-lvm'})
|
||||
|
||||
def days_pass(self, days):
|
||||
"""Moves every date of the record back, as if that time had gone by."""
|
||||
conn = self.store._get_conn()
|
||||
row = conn.execute('SELECT first_seen, last_seen, acknowledged_at, resolved_at FROM errors WHERE error_key = ?',
|
||||
(KEY,)).fetchone()
|
||||
if row:
|
||||
moved = [(datetime.fromisoformat(v) - timedelta(days=days)).isoformat() if v else None for v in row]
|
||||
conn.execute('UPDATE errors SET first_seen = ?, last_seen = ?, acknowledged_at = ?, resolved_at = ? '
|
||||
'WHERE error_key = ?', (*moved, KEY))
|
||||
conn.commit()
|
||||
conn.close()
|
||||
|
||||
def cleanup(self):
|
||||
with patch('builtins.print'), patch('subprocess.run', side_effect=OSError('no host tools in the test')):
|
||||
self.store.cleanup_old_errors()
|
||||
|
||||
def test_it_is_still_dismissed_weeks_later(self):
|
||||
self.assertTrue(self.seen()['needs_notification'])
|
||||
self.store.acknowledge_error(KEY, suppression_hours=-1)
|
||||
for _ in range(6):
|
||||
self.days_pass(8)
|
||||
self.cleanup()
|
||||
self.assertFalse(self.seen()['needs_notification'])
|
||||
self.assertTrue(self.store.is_error_permanently_acknowledged(KEY))
|
||||
|
||||
def test_a_condition_that_recovers_and_returns_stays_dismissed(self):
|
||||
self.seen()
|
||||
self.store.acknowledge_error(KEY, suppression_hours=-1)
|
||||
self.store.clear_error(KEY)
|
||||
self.store.resolve_error(KEY, 'condition recovered')
|
||||
self.days_pass(8)
|
||||
self.cleanup()
|
||||
self.assertFalse(self.seen()['needs_notification'])
|
||||
self.assertTrue(self.store.is_error_permanently_acknowledged(KEY))
|
||||
|
||||
def test_a_record_already_closed_for_its_age_is_kept(self):
|
||||
self.seen()
|
||||
self.store.acknowledge_error(KEY, suppression_hours=-1)
|
||||
conn = self.store._get_conn()
|
||||
closed = (datetime.now() - timedelta(days=8)).isoformat()
|
||||
conn.execute("UPDATE errors SET resolved_at = ?, resolution_type = 'auto', "
|
||||
"resolution_reason = 'Stale error (no activity >7d)' WHERE error_key = ?", (closed, KEY))
|
||||
conn.commit()
|
||||
conn.close()
|
||||
self.cleanup()
|
||||
self.assertFalse(self.seen()['needs_notification'])
|
||||
|
||||
def test_saving_the_suppression_settings_leaves_it_permanent(self):
|
||||
self.seen()
|
||||
self.store.acknowledge_error(KEY, suppression_hours=-1)
|
||||
with patch('builtins.print'):
|
||||
self.store.sync_dismissed_suppression()
|
||||
self.assertTrue(self.store.is_error_permanently_acknowledged(KEY))
|
||||
|
||||
def test_re_enabling_it_brings_the_alert_back(self):
|
||||
self.seen()
|
||||
self.store.acknowledge_error(KEY, suppression_hours=-1)
|
||||
self.store.unacknowledge_error(KEY)
|
||||
self.assertFalse(self.store.is_error_permanently_acknowledged(KEY))
|
||||
|
||||
def test_a_timed_dismissal_still_ends(self):
|
||||
self.seen()
|
||||
self.store.acknowledge_error(KEY, suppression_hours=24)
|
||||
self.days_pass(2)
|
||||
self.assertTrue(self.seen()['needs_notification'])
|
||||
|
||||
|
||||
if __name__ == '__main__':
|
||||
unittest.main()
|
||||
Reference in New Issue
Block a user