Policies & Governance

Moonin governance is built around three operational policy families:

• alert policies
• event notification policies
• scaling rules

This page documents the behavior behind:

• https://app.moonin.app/alert-policies
• https://app.moonin.app/event-notification-policies
• https://app.moonin.app/scaling-rules

Policy hierarchy

All three policy families inherit the organization boundary and then narrow the target scope through workload path information.

flowchart TD
    O[Organization]
    S[Scope filters]
    C[Channels or actions]
    R[Runtime result]

    O --> S --> C --> R

Alert policies

Alert policies are error-driven. They evaluate captured deployment revision errors and decide whether a notification should be sent.

Alert policy inputs

An alert policy can combine:

• name and description
• enabled state
• one or more error types
• minimum affected ratio
• path matching
• project, cluster, namespace and deployment scope
• silence window
• delay in seconds
• one or more notification channels
• active UTC windows per channel

How alert matching works

flowchart LR
    A[Revision error captured]
    B[Scope and path match]
    C[Error type match]
    D[Affected ratio match]
    E[Delay elapsed]
    F[Policy not silenced]
    G[Active channel window]
    H[Send alert]

    A --> B --> C --> D --> E --> F --> G --> H

What the delay means

The delay prevents immediate alerting on very fresh failures. Moonin waits until the configured number of seconds has elapsed since the error occurrence before sending.

Use delay when:

• the error is often transient during startup
• you want to reduce noise from short-lived rollout turbulence

What the affected ratio means

The affected ratio reflects how much of the workload is impacted. This lets teams distinguish:

• one unhealthy pod in a larger rollout
• a broad service failure affecting most or all pods

Event notification policies

Event notification policies are not failure-driven. They route lifecycle events to the right channels.

Event policy inputs

An event notification policy can combine:

• name and description
• enabled state
• event types
• path matching
• silence window
• one or more channels
• active UTC windows per channel

Current event model

The current documented event flow includes:

• deployment.revision.created

This means event policies are useful when teams want deployment awareness even before incidents exist.

Event evaluation flow

flowchart LR
    A[Revision created]
    B[Event policy path match]
    C[Event type match]
    D[Policy not silenced]
    E[Channel active now]
    F[Send event notification]

    A --> B --> C --> D --> E --> F

Scaling rules

Scaling rules are scheduled or manual runtime actions designed to temporarily override service scaling behavior in a controlled and auditable way.

What a scaling rule contains

• name
• description
• cron expression
• timezone
• optional duration in minutes
• optional expiration date
• enabled state
• creator information
• one or more scaling actions attached to deployments

What a scaling action contains

For each selected deployment, Moonin can store:

• project
• cluster
• namespace
• deployment
• minimum replicas
• maximum replicas
• default replicas
• HPA context when present

Scaling rule flow

flowchart LR
    A[Create rule template]
    B[Attach deployment actions]
    C[Rule becomes eligible by schedule]
    D[Manual or scheduled execution runs]
    E[Temporary scaling window applies]
    F[Audit and events recorded]

    A --> B --> C --> D --> E --> F

Duration and expiration behavior

• duration_minutes controls how long a rule execution should remain active
• valid_until defines the last date at which the rule is considered valid
• a rule can be scheduled or run manually
• manual execution can be created and cancelled independently of the template definition

What operators typically do with scaling rules

• prepare for traffic peaks
• reduce capacity after a known busy window
• coordinate temporary scaling during maintenance or releases
• document which deployments are being intentionally scaled and by how much

Auditability

Scaling rules include supporting views for:

• action review
• event history
• audit history
• manual execution state

Treat these screens as the operational source of truth for planned temporary scaling.

Choosing the right policy family

Use Alert Policies when the trigger is a failure.

Use Event Notification Policies when the trigger is an operational event, such as a new revision.

Use Scaling Rules when the desired outcome is a temporary runtime scaling change rather than a notification.