Advanced Istio Best Practices – Locality Routing and Service Mesh Optimization

This article, authored by delphisfang from Tencent Music, presents a personal best‑practice guide for advancing Istio usage, targeting developers who want to move beyond introductory tutorials and quickly master locality‑based routing in a Service Mesh.

The author introduces a "Three‑Evidence Learning Method" (official documents, source code, and prototypes) to systematically explore complex systems. By asking concrete questions, researching documentation, building prototypes, and digging into the source, readers can efficiently acquire deep knowledge.

Locality in cloud environments is defined as a three‑tuple <Region, Zone, Sub‑zone> . Istio extends this concept to service instances, allowing traffic to be preferentially routed to the nearest instance based on matching locality levels.

The core algorithm, implemented in applyLocalityFailover (Istio source), assigns a priority to each endpoint:

If region, zone, and sub‑zone all match, priority = 0.

If only region and zone match, priority = 1.

If only region matches, priority = 2.

If no match, priority = 3.

Envoy does not compute distances itself; Istio provides the priority via the locality and priority fields in the endpoint configuration. The article includes a concrete Envoy endpoint configuration example:

- name: helloworld
  connect_timeout: 0.1s
  type: static
  lb_policy: ROUND_ROBIN
  load_assignment:
    cluster_name: helloworld
    endpoints:
    - priority: 0
      locality:
        region: south-china
        zone: shenzhen
        sub_zone: 01
      lb_endpoints:
      - endpoint:
          address:
            socket_address:
              address: 1.1.1.1
              port_value: 12345
    - priority: 1
      locality:
        region: south-china
        zone: shenzhen
        sub_zone: 02
      lb_endpoints:
      - endpoint:
          address:
            socket_address:
              address: 2.2.2.2
              port_value: 12345
    - priority: 2
      locality:
        region: south-china
        zone: guangzhou
        sub_zone: 01
      lb_endpoints:
      - endpoint:
          address:
            socket_address:
              address: 3.3.3.3
              port_value: 12345
    - priority: 3
      locality:
        region: east-china
        zone: shanghai
        sub_zone: 01
      lb_endpoints:
      - endpoint:
          address:
            socket_address:
              address: 4.4.4.4
              port_value: 12345

To enable locality‑aware load balancing, Istio requires a DestinationRule with localityLbSetting and an outlierDetection block. A minimal example is provided:

apiVersion: networking.istio.io/v1beta1
kind: DestinationRule
metadata:
  name: helloworld
  namespace: demo-namespace
spec:
  host: helloworld.svc.cluster.local
  trafficPolicy:
    connectionPool:
      http:
        maxRequestsPerConnection: 100
    loadBalancer:
      simple: ROUND_ROBIN
    localityLbSetting:
      enabled: true
      failover: []
    outlierDetection:
      consecutive5xxErrors: 10
      interval: 5s
      baseEjectionTime: 1m

The article also explains how Istio discovers the locality of each Envoy instance via the Proxy object populated from the Envoy bootstrap configuration, and how ServiceEntry resources embed locality information for each endpoint.

Finally, the author outlines a simplified Istio architecture focusing on the essential components (Pilot and Envoy) and provides practical advice for reducing complexity during learning and experimentation.