Docsright arrowEmissary-ingressright arrowThe Emissary-ingress architecture

2 min • read

The Emissary-ingress architecture

Emissary-ingress is a control plane

Emissary-ingress is a specialized control plane for Envoy Proxy. In this architecture, Emissary-ingress translates configuration (in the form of Kubernetes Custom Resources) to Envoy configuration. All actual traffic is directly handled by the high-performance Envoy Proxy.

Architecture

Details

  1. The service owner defines configuration in Kubernetes manifests.
  2. When the manifest is applied to the cluster, the Kubernetes API notifies Emissary-ingress of the change.
  3. Emissary-ingress parses the change and transforms the configuration into a semantic intermediate representation. Envoy configuration is generated from this IR.
  4. The new configuration is passed to Envoy via the gRPC-based Aggregated Discovery Service (ADS) API.
  5. Traffic flows through the reconfigured Envoy, without dropping any connections.

Scaling and availability

Emissary-ingress relies on Kubernetes for scaling, high availability, and persistence. All Emissary-ingress configuration is stored directly in Kubernetes; there is no database. Emissary-ingress is packaged as a single container that contains both the control plane and an Envoy Proxy instance. By default, Emissary-ingress is deployed as a Kubernetes deployment and can be scaled and managed like any other Kubernetes deployment.

Stateless architecture

By design, Emissary-ingress is an entirely stateless architecture. Each individual Emissary-ingress instance operates independently of other instances. These Emissary-ingress instances rely on Kubernetes to coordinate the configuration between the different Emissary-ingress instances. This enables Emissary-ingress to sidestep the need to engineer a safe, highly available centralized control plane (and if you don't think that this is hard, check out Jepsen). By contrast, other control plane architectures rely on a single centralized control plane to manage multiple instances of the data plane. This means that these control plane architectures must engineer resilience and availability into their central control plane.

Envoy Proxy

Emissary-ingress closely tracks Envoy Proxy releases. A stable branch of Envoy Proxy is maintained that enables the team to cherry-pick specific fixes into Emissary-ingress.