Google Cloud Platform Spanner

Scale applications based on the result of a Google Cloud Spanner SQL query.

Availability: v2.21+ Maintainer: Community

Trigger Specification

This specification describes the gcp-spanner trigger for Google Cloud Spanner.

The scaler executes a user-supplied SQL statement against a Spanner database on every polling interval. The query must return a single row whose first column is an INT64 value — typically a COUNT(*) or SUM(...) expression that represents current workload depth.

triggers:
- type: gcp-spanner
  metadata:
    projectId: my-gcp-project          # Required
    instanceId: my-spanner-instance    # Required
    databaseId: my-database            # Required
    query: "SELECT COUNT(*) FROM jobs WHERE status = 'pending'"  # Required
    targetValue: "5"                   # Optional - default: 5
    activationValue: "0"               # Optional - default: 0
    credentialsFromEnv: GOOGLE_APPLICATION_CREDENTIALS_JSON  # Required*

* One of credentialsFromEnv, credentialsFromEnvFile, or a TriggerAuthentication referencing GoogleApplicationCredentials / GCP pod identity is required (unless SPANNER_EMULATOR_HOST is set, in which case authentication is handled by the emulator).

Parameter list:

projectId - The GCP project that owns the Spanner instance.
instanceId - The Spanner instance ID.
databaseId - The Spanner database ID.
query - A SQL statement that returns exactly one row with one INT64 column. When the query matches no rows the value is treated as 0 (not an error), so both COUNT(*) aggregates and single-value SELECT statements work correctly.
targetValue - The value of the metric per replica. KEDA and the HPA use ceil(currentValue / targetValue) to compute the desired replica count. (Default: 5, Optional)
activationValue - Minimum metric value required to activate the scaler. While currentValue ≤ activationValue the scaler is considered inactive and the deployment is held at 0 replicas. Learn more about activation here. (Default: 0, Optional)
credentialsFromEnv - Name of an environment variable on the scale target that contains the service-account JSON.
credentialsFromEnvFile - Name of an environment variable on the scale target whose value is a path to a JSON key file.

The metric name is generated automatically from the trigger index and the resource identifiers, for example: s0-gcp-spanner-<instanceId>-<databaseId>-<projectId>.

Authentication Parameters

You can use a TriggerAuthentication CRD to supply credentials outside of the scale-target environment.

Credential-based authentication (service account JSON):

GoogleApplicationCredentials - Service account credentials in JSON.

Identity-based authentication (Workload Identity / GKE metadata server):

Create a TriggerAuthentication with podIdentity.provider: gcp and reference it from the ScaledObject. No JSON key is required.

Examples

Inline credentials via environment variable

apiVersion: keda.sh/v1alpha1
kind: ScaledObject
metadata:
  name: spanner-scaledobject
  namespace: my-namespace
spec:
  scaleTargetRef:
    name: job-processor
  pollingInterval: 15
  cooldownPeriod: 30
  minReplicaCount: 0
  maxReplicaCount: 10
  triggers:
    - type: gcp-spanner
      metadata:
        projectId: my-gcp-project
        instanceId: my-spanner-instance
        databaseId: my-database
        query: "SELECT COUNT(*) FROM jobs WHERE status = 'pending'"
        targetValue: "5"
        activationValue: "2"
        credentialsFromEnv: GOOGLE_APPLICATION_CREDENTIALS_JSON

Using TriggerAuthentication with a Kubernetes Secret

apiVersion: v1
kind: Secret
metadata:
  name: spanner-secret
  namespace: my-namespace
stringData:
  creds.json: |
    { "type": "service_account", ... }
---
apiVersion: keda.sh/v1alpha1
kind: TriggerAuthentication
metadata:
  name: keda-trigger-auth-gcp-spanner
  namespace: my-namespace
spec:
  secretTargetRef:
    - parameter: GoogleApplicationCredentials
      name: spanner-secret
      key: creds.json
---
apiVersion: keda.sh/v1alpha1
kind: ScaledObject
metadata:
  name: spanner-scaledobject
  namespace: my-namespace
spec:
  scaleTargetRef:
    name: job-processor
  triggers:
    - type: gcp-spanner
      authenticationRef:
        name: keda-trigger-auth-gcp-spanner
      metadata:
        projectId: my-gcp-project
        instanceId: my-spanner-instance
        databaseId: my-database
        query: "SELECT COUNT(*) FROM jobs WHERE status = 'pending'"
        targetValue: "5"
        activationValue: "2"

Using TriggerAuthentication with GCP Workload Identity

apiVersion: keda.sh/v1alpha1
kind: TriggerAuthentication
metadata:
  name: keda-trigger-auth-gcp-spanner
  namespace: my-namespace
spec:
  podIdentity:
    provider: gcp
---
apiVersion: keda.sh/v1alpha1
kind: ScaledObject
metadata:
  name: spanner-scaledobject
  namespace: my-namespace
spec:
  scaleTargetRef:
    name: job-processor
  triggers:
    - type: gcp-spanner
      authenticationRef:
        name: keda-trigger-auth-gcp-spanner
      metadata:
        projectId: my-gcp-project
        instanceId: my-spanner-instance
        databaseId: my-database
        query: "SELECT COUNT(*) FROM jobs WHERE status = 'pending'"
        targetValue: "5"