Skip to content
Phone Guides

Asterisk monitoring and alert design

Monitor the service people need, then the components that explain its failure. An Asterisk process responding to a metrics request does not establish that a phone can receive an ordinary call, that audio works or that voicemail can store a message.

Source-reviewed, 18 October 2026. Prepared from the cited documentation. Monitoring integrations and alert thresholds have not been independently tested by Phone Guides. Scope: Asterisk 22 operational design; choose collectors and metrics supported by the actual build.

Build a small service map

Record the Asterisk host, endpoints, provider, DNS, network edge, certificates, storage, databases and integration applications needed by each calling path. Identify an owner and recovery action for each dependency. Keep separate sites and trunks distinguishable in the alert.

Check Example operational question
Process/service state Did the intended instance start and stay running?
Registration/contact state Is the required peer relationship present?
Calling path Can the intended authorised destination ring, answer and clear?
Media Does the required audio/DTMF path behave correctly?
Storage/backend Can required logs, messages and recordings be written?
Certificate/trust Is renewal due, and does the protected connection still validate?
Integration dependency Is the database/API available within the call's allowed delay?

Registration and reachability checks have limits. Record what a contact or qualification result means for that specific peer rather than displaying a single green indicator for every service.

Select a protected collection path

Use documented CLI, logs, events or metrics according to the information needed. If choosing Asterisk's Prometheus module, consult the branch-22 res_prometheus reference. Confirm its required service configuration and authentication, and restrict access to approved collectors.

The module reference warns about exposing system information through an unauthenticated metrics route. Treat metric access as administrative information access, not a reason to publish a new listener to the whole Internet. Individual metrics/options can also be added in later patches within a branch; check their Since fields before using them.

Define alerts with a response

For each alert, write the observed condition, scope, threshold/duration, recovery owner and first safe diagnostic step. Choose thresholds from the service's required behaviour and measured history when available. This guide supplies no universal CPU, delay or call-count threshold.

Avoid repeated notifications for the same unchanged incident. Group related failures so a lost network path does not generate an unrelated-looking alert for every endpoint. Include recovery events and planned maintenance treatment so operators know when the state changes.

Keep synthetic calls bounded

If your organisation adds test calls, use authorised ordinary destinations and a stated frequency, cost limit and cleanup procedure. Do not call real emergency numbers as a monitoring check. Establish who answers or verifies the media and how an abandoned test call is cleared.

A call that rings proves a different stage from an answered two-way-audio check. State which stage the monitor observes. No synthetic calling is required to publish this source-reviewed guide, and Phone Guides has not run a test call here.

Preserve useful incident evidence

Connect an alert with its time/timezone, service scope, recent change and relevant identifiers. Use logs, CDR, CEL and queue_log according to the question. Restrict subscriber data and credentials, and apply a suitable retention policy.

For your own monitoring checks, simulate an appropriate controlled dependency failure and verify detection, notification and recovery. A collector receiving data is only the first part of that check. Review first troubleshooting checks and recovery planning for the next actions.

Return to Asterisk documentation.