Metric Trigger Rules

The Coordimap agent can evaluate metric rules directly and send metric-trigger elements to the backend. These are sent as regular elements with the type coordimap.metric_trigger and include all matching internal IDs in the element payload, automatically relating alerts to your infrastructure graph.

Metric rules are configured inside each datasource block under data_sources[*].metric_rules.

Currently, metric rules are supported only for the following datasource types:

• kubernetes
• gcp

Supported Providers

• Prometheus: For Kubernetes data sources.
• GCP Monitoring: For Google Cloud Platform (GCP) data sources.

Configuration Format

Each rule must specify a mode, which can be either:

• custom: For user-defined provider queries.
• predefined: For built-in templates.

Example Configuration

coordimap:
  data_sources:
    - type: kubernetes
      id: kube-prod
      config:
        - name: scope_id
          value: your-cluster-uid
        - name: config_file
          value: /path/to/your/kube/config
      metric_rules:
        # Example of a custom Prometheus query
        - id: k8s-high-5xx
          name: Kubernetes Service High 5xx
          provider: prometheus
          mode: custom
          custom:
            query: sum(rate(istio_requests_total{response_code=~"5.."}[5m])) by (destination_workload_namespace, destination_canonical_service)
          lookback: 5m
          threshold:
            operator: ">"
            value: 1
          target:
            resolver: kubernetes_service
            namespace_label: destination_workload_namespace
            name_label: destination_canonical_service

    - type: gcp
      id: gcp-prod
      config:
        - name: scope_id
          value: "123456789012"
        - name: project_id
          value: your-project-id
      metric_rules:
        # Example of a predefined GCP Monitoring template
        - id: cloudsql-high-cpu
          name: CloudSQL High CPU
          provider: gcp_monitoring
          mode: predefined
          predefined:
            name: cloudsql_high_cpu
            params:
              lookback: 5m
              threshold: 0.8

Field Reference

Common Fields

All metric rules share these fields:

• id: A unique identifier for this rule.
• name: A human-readable name for the trigger.
• provider: The monitoring backend (prometheus or gcp_monitoring).
• lookback: The time window to evaluate (e.g., 5m).
• threshold.operator and threshold.value: Conditions for triggering the rule (e.g., > and 1).
• target.resolver: The asset type this rule applies to.

Prometheus-specific Fields

When using mode: custom with the prometheus provider, you must specify:

• custom.query: The PromQL query to evaluate.

GCP Monitoring-specific Fields

When using mode: custom with the gcp_monitoring provider, you must specify:

• custom.filter or custom.metric_type: To select the correct GCP metric.
• Optional fields: alignment_period, per_series_aligner, cross_series_reducer, and group_by_fields to control aggregation.

Predefined Rules

Predefined rules allow you to use built-in templates without writing custom queries.

Prometheus Templates

• kubernetes_service_high_5xx (Params: window, threshold)
• kubernetes_deployment_high_5xx (Params: window, threshold)
• kubernetes_service_high_latency (Params: window, quantile, threshold)
• kubernetes_pod_high_restart_rate (Params: window, threshold)
• kubernetes_pod_crashloop_or_imagepull_error (Params: lookback, reason_regex)
• kubernetes_pod_not_ready (Params: lookback)
• kubernetes_deployment_unavailable_replicas (Params: threshold)
• kubernetes_deployment_availability_gap (Params: threshold)
• kubernetes_pod_high_cpu_usage (Params: window, threshold)
• kubernetes_pod_high_memory_workingset (Params: threshold)
• kubernetes_pod_cpu_throttling_high (Params: window, threshold)
• kubernetes_pod_oom_events (Params: window)
• kubernetes_pod_unschedulable (Params: lookback)
• kubernetes_pvc_low_free_space (Params: threshold, namespace, pvc_regex)
• kubernetes_pvc_free_space_burn_rate (Params: threshold, window, horizon_seconds, namespace, pvc_regex)
• kubernetes_inode_low_free (Params: threshold, namespace, pvc_regex)
• kubernetes_statefulset_pvc_low_free_space (Params: threshold, namespace, statefulset, volume_claim_prefix)

GCP Monitoring Templates

• cloudsql_high_cpu (Params: lookback, threshold)
• cloudsql_high_connections (Params: lookback, threshold, metric_type, alignment_period, per_series_aligner)
• vm_high_cpu (Params: lookback, threshold, alignment_period, per_series_aligner)

Target Resolvers

The target.resolver field maps the results of a metric query to specific asset identities within the Coordimap graph.

Available resolvers include:

• Kubernetes: kubernetes_service, kubernetes_deployment, kubernetes_pod, kubernetes_pvc, kubernetes_statefulset
• GCP: gcp_cloudsql, gcp_vm_instance
• Cross data source: external_mapping

When using the external_mapping resolver, if no external_mappings entry matches the emitted target, the target is ignored and nothing is sent for that series.