ENVOY

Envoy and service mesh data plane engineering

RTZ Labs works on Envoy proxy and service mesh data plane behavior – circuit breaking, connection pooling, load balancing, and failure isolation – for teams running Envoy directly or through Istio in production.

Envoy is the component actually carrying your traffic, whether you configure it directly or through a mesh control plane. Most mesh behavior that surprises people is Envoy behavior: connection pool limits applied per-host rather than per-service, outlier detection ejecting more hosts than expected, or circuit breakers that were never reached because the thresholds were set far above real concurrency. We read the proxy statistics and work from what the data plane is doing.

All expertise

What we do with Envoy

  • Set circuit breaker budgets – max connections, pending requests, and retries – against measured concurrency rather than defaults
  • Diagnose connection pool exhaustion and the upstream saturation it hides
  • Review load balancing policy and locality-aware routing, including its interaction with autoscaling
  • Analyze outlier detection and host ejection behavior during partial failures
  • Interpret Envoy statistics and access logs to locate where a request actually spends its time
  • Review timeout and retry budgets across service hops so retries cannot multiply load

Problems we are called in for

Usually described this way before anyone knows the cause.

Upstream services saturate under load while the circuit breaker never trips
Traffic distributes unevenly across replicas despite a working autoscaler
Requests fail with upstream connect errors that do not match the backend health
Retry storms turn a slow dependency into an outage
Envoy metrics are being collected but nobody can act on them

Scope and boundaries

We work on Envoy as deployed in Kubernetes – as an Istio data plane, as an ingress or gateway, or standalone. We do not develop custom Envoy C++ filters.

Provider
RTZ Labs
Availability
Remote, serving the United States and Canada
How to start
Email contact@rtzlabs.io or use the contact form.

Production Reliability

Reduce incidents, improve observability, and build infrastructure your team can trust – on AWS and Kubernetes.

Cloud Platform Engineering

Build AWS and Kubernetes platforms that let developers ship safely and consistently – IaC, GitOps, CI/CD, and networking.