FAQ
What changed from the previous Mimir-based alerting?
The advanced alerting system uses the full Grafana Ruler and Alertmanager stack. The key differences are:
- Folders are required. Rules must be organised into folders.
- Evaluation groups are required. The evaluation interval is set at the group level, not per rule.
- Routing uses labels. Alerts are routed to contact points via notification policies and label matchers, rather than by direct assignment at the rule level.
- More configuration options are available, including multi-source queries, math expressions, and custom state handling.
See Configure Alert Rules for a full walkthrough of the new setup.
What is the difference between a Contact Point and a Notification Policy?
A contact point defines where and how a notification is sent (such as, a specific Slack channel or email address).
A notification policy defines which alerts are routed to which contact point, based on the labels on the alert rule.
Think of it as: the notification policy is the routing rule, and the contact point is the destination.
How do I route an alert to a specific contact point?
Add a channel label to your alert rule (such as, channel=slack) and ensure a notification policy exists with a matching label matcher (such as, channel =~ .*slack.*).
See Notification Policies for the recommended setup.
When should I use a Silence vs. a Mute Timing?
| Silence | Mute Timing (Time Interval) | |
|---|---|---|
| Use for | One-off suppression (such as, an emergency maintenance window) | Recurring schedules (such as, every weekend or overnight) |
| Set up via | Alerting > Silences | Alerting > Notification policies > Time intervals |
| Applied to | Individual alerts matched by label | Notification policies |
See Silences and Time Intervals for setup instructions.
Can I route the same alert to both Slack and email?
Yes. There are two ways:
-
Single contact point with multiple integrations — Add both a Slack integration and an email integration to the same contact point. Both will receive every notification sent to that contact point.
-
Multiple policies with Continue matching — Create separate child policies for each destination and enable Continue matching on each one. An alert matching the first policy will continue to be evaluated against sibling policies.
How do I test whether my labels will route correctly?
Use the Test option on the Default policy in Alerting > Notification policies. Enter the label key/value pairs from your alert rule to simulate routing and see which policy and contact point the alert would be delivered to.
You can also click Test on an individual contact point to verify that the integration credentials and configuration are working correctly.
My alert fired but I did not receive a runbook link in the notification
Runbook links are set as an annotation on the alert rule. Edit the rule, scroll to the Configure notification message section, and add a Runbook URL value.
What does the Pending period do?
The pending period is how long the alert condition must be continuously met before the alert transitions from Pending to Firing. This prevents notifications for brief, transient spikes.
Setting it to 0s means the alert fires immediately on the first evaluation where the condition is true. A value of 5m means the condition must hold for five consecutive minutes before a notification is sent.
Why is my alert in No Data state?
No Data means the alert rule's query returned no results during evaluation. Common causes:
- The FusionReactor agent is not sending data for the monitored instance.
- The query label selectors are too specific and match no active series.
- The data source connection is temporarily unavailable.
Review the No Data handling option on the rule — Keep last state is usually the safest default. See Troubleshooting for more detail.
How do I prevent an alert from repeatedly firing and recovering?
Enable Keep firing for on the alert rule. This holds the alert in a firing state for a defined period after the condition resolves, preventing noisy recovered/re-fired cycles for metrics that fluctuate around the threshold.
Increasing the Pending period also helps — a longer pending period requires the condition to be stable before firing.
What happens to alerts that don't match any notification policy?
They are handled by the Default policy, which acts as a safety net. Configure the Default policy to point to a catch-all contact point (such as, a general-purpose email or Slack channel) to ensure no alert is silently dropped.