Scale Management

1. Background

Scale management is used to adjust the replica count of the ShenYu Bootstrap deployment in Kubernetes. ShenYu Admin can scale by fixed time policies or by dynamic metric rules collected from Prometheus.

This feature is disabled by default. Enable it only when ShenYu Admin can access the Kubernetes API server and has permission to update the target Deployment scale subresource.

2. Configuration

Enable scale management in application.yml:

shenyu:
  k8s:
    scale:
      enabled: true
      monitor-interval: 10000
      pool-size: 6
    prometheus:
      url: http://localhost:9090
      queries:
        cpu_usage: "sum(rate(container_cpu_usage_seconds_total{namespace='%s', pod=~'%s.*'}[5m]))"
        memory_usage: "sum(container_memory_usage_bytes{namespace='%s', pod=~'%s.*'})"
        request_count: "sum(rate(http_requests_total{namespace='%s', pod=~'%s.*'}[1m]))"
    deployment:
      name: "shenyu-bootstrap"
      namespace: "shenyu"
      apiServer: "https://127.0.0.1:6443"
      token: "token"
      caCertPath: "/etc/kubernetes/pki/ca.crt"

Configuration description:

Name	Description
shenyu.k8s.scale.enabled	Whether to enable Kubernetes scale management. The default value is false.
shenyu.k8s.scale.monitor-interval	Metric polling interval in milliseconds.
shenyu.k8s.scale.pool-size	Thread pool size for scale monitoring tasks.
shenyu.k8s.prometheus.url	Prometheus server URL.
shenyu.k8s.prometheus.queries	PromQL templates used by metric rules.
shenyu.k8s.deployment.name	Target Kubernetes Deployment name.
shenyu.k8s.deployment.namespace	Namespace of the target Deployment.
shenyu.k8s.deployment.apiServer	Kubernetes API server address.
shenyu.k8s.deployment.token	Service account token used to access the Kubernetes API server.
shenyu.k8s.deployment.caCertPath	CA certificate path for the Kubernetes API server.

3. Fixed Scale Policies

Choose System Management -> Scale Management in ShenYu Admin.

Scale policies define a fixed number of Bootstrap replicas within a time range. The policy fields include:

Field	Description
Sort	Policy priority.
Status	1 means enabled, 0 means disabled.
Num	Target number of Bootstrap replicas.
Begin Time	The time when the policy starts to take effect.
End Time	The time when the policy stops taking effect.

When an enabled policy is active, ShenYu Admin scales the target Deployment to the configured replica count.

4. Dynamic Scale Rules

Scale rules define upper and lower thresholds for metrics. The rule fields include:

Field	Description
Metric Name	Metric key. It must match a key under shenyu.k8s.prometheus.queries.
Type	Metric source type. 0 is ShenYu, 1 is Kubernetes, and 2 is others.
Sort	Rule priority.
Status	1 means enabled, 0 means disabled.
Minimum	If the collected metric value is lower than this value, ShenYu Admin scales down by one replica.
Maximum	If the collected metric value is higher than this value, ShenYu Admin scales up by one replica.

Dynamic scaling uses the Prometheus query configured for the metric name. The query template receives the target Kubernetes namespace and Deployment name.

5. Notes

• Configure Kubernetes RBAC before enabling scale management. The token must be allowed to read and update the target Deployment scale.
• Keep the metric name in scale rules consistent with the keys configured under shenyu.k8s.prometheus.queries.
• Start with conservative thresholds and verify the scale history before using the feature in production.