Create an Alert Rule
Start by defining the service impact, first owner, urgency, and expected response action. If a responder cannot act on the notification, improve the signal before creating the alert.
Grafana Alert
Step 1: Define the Query
Develop PromQL over representative healthy, degraded, and missing-data windows. Preserve labels needed for customer/device ownership and routing.
Step 2: Set Conditions
Choose an evaluation interval, pending duration, threshold, no-data behavior, and evaluation-error behavior that detect sustained impact without creating transient noise.
Step 3: Configure Notifications
Add severity, owner, runbook/dashboard links, and the correct contact-point label. Confirm grouping does not combine unrelated customers.
Step 4: Test
Test firing, pending-to-alerting transition, notification delivery, acknowledgement, and resolved notification. Deploy through the managed alert repository.
Splunk Alert
Step 1: Define the Search
Use an explicit index and bounded time window. Confirm all selected sourcetypes provide the fields required for routing and payloads.
Step 2: Set Conditions
Define schedule, trigger type, comparator, threshold, severity, suppression fields/period, and per-result versus digest behavior.
Step 3: Configure Notifications
Select the approved action/contact point and remove secrets or customer-sensitive content from the payload.
Create and deploy the alert through the managed Terraform repository; avoid untracked UI-only production alerts.