Skip to main content

Alert Rules

Configure monitoring rules under "Alert Management → Alert Rules". Select built-in metrics or use custom PromQL expressions.

Set thresholds and notification policies according to your environment. Improper configuration may cause false positives or missed alerts. Validation in a test environment before production use is recommended.

Alert rules

Monitoring Types​

Rules are grouped by monitoring type. Middleware types appear only after the corresponding Data Source is enabled:

TabMonitoring ScopeCondition
Basic ResourcesCPU, memory, disk, and other host/middleware metrics; custom PromQL is supportedAlways
PortTCP port connectivity probingAlways
SSLHTTPS certificate validity periodAlways
MySQLSpecialized checks such as connections and slow queriesMySQL data source enabled
MongoDBSpecialized checks such as connections and replication lagMongoDB data source enabled
Kafka LagConsumer group lag backlogKafka data source enabled
Kafka RebalanceFrequent consumer group rebalanceKafka data source enabled

Create a Rule​

Click "New Task" and fill in the fields:

FieldDescription
Task NameUnique and readable, such as High production host CPU
Data SourceTarget instance to monitor
MetricSelect a built-in metric, or choose "Custom PromQL" to enter an expression and preview the results
Alert ThresholdOperator + threshold, such as > 80
Check FrequencyHow often the rule is checked
DurationHow long the metric must continuously exceed the threshold before entering the alert state, filtering out short spikes
Notification ChannelBind a notification channel; if not bound, records are kept only inside the Ops Platform

State Flow​

Rules send notifications when their state changes. A continuously firing alert does not send repeated notifications:

Normal ──threshold exceeded──▶ Pending ──duration satisfied──▶ Firing ──recovered──▶ Recovered
StateMeaning
PendingThreshold exceeded, but duration not yet satisfied
FiringThreshold exceeded and duration satisfied; notification has been sent
RecoveredRecovered from alert state; recovery notification sent
PausedRule manually paused and no longer scheduled

Common Operations​

  • Check Now: Skip the wait and run one check immediately.
  • Pause/Enable: Temporarily stop scheduling during planned maintenance, replacing the old "alert silence" flow.
  • Instance Details: View current values for matched instances. Per-instance silence is supported, so individual instances can be suppressed while the rule continues to run.
  • Alert History: View trigger records for this rule. See Alert History for records from all rules.