Permalink
Switch branches/tags
Nothing to show
Find file Copy path
34f3b9d Apr 16, 2018
4 contributors

Users who have contributed to this file

@htuch @mattklein123 @alyssawilk @snowp
129 lines (94 sloc) 7.18 KB

Envoy v2 APIs for developers

Goals

This repository contains both the implemented and draft v2 JSON REST and gRPC Envoy APIs.

Version 2 of the Envoy API evolves existing APIs and introduces new APIs to:

  • Allow for more advanced load balancing through load and resource utilization reporting to management servers.
  • Improve N^2 health check scalability issues by optionally offloading health checking to other Envoy instances.
  • Support Envoy deployment in edge, sidecar and middle proxy deployment models via changes to the listener model and CDS/SDS APIs.
  • Allow streaming updates from the management server on change, instead of polling APIs from Envoy. gRPC APIs will be supported alongside JSON REST APIs to provide for this.
  • Ensure all Envoy runtime configuration is dynamically discoverable via API calls, including listener configuration, certificates and runtime settings, which are today sourced from the filesystem. There will still remain a static bootstrap configuration file that will specify items unlikely to change during runtime, including the Envoy node identity, xDS management server addresses, administration interface and tracing configuration.
  • Revisit and where appropriate cleanup any v1 technical debt.

Status

See here for the current status of the v2 APIs.

See here for the v2 API change process.

Principles

  • Proto3 will be used to specify the canonical API. This will provide directly the gRPC API and via gRPC-JSON transcoding the JSON REST API. A textual YAML input will be supported for filesystem configuration files (e.g. the bootstrap file), in addition to JSON, as a syntactic convenience. YAML file contents will be internally converted to JSON and then follow the standard JSON-proto3 conversion during Envoy config ingestion.

  • xDS APIs should support eventual consistency. For example, if RDS references a cluster that has not yet been supplied by CDS, it should be silently ignored and traffic not forwarded until the CDS update occurs. Stronger consistency guarantees are possible if the management server is able to sequence the xDS APIs carefully (for example by using the ADS API below). By following the [CDS, EDS, LDS, RDS] sequence for all pertinent resources, it will be possible to avoid traffic outages during configuration update.

  • The API is primarily intended for machine generation and consumption. It is expected that the management server is responsible for mapping higher level configuration concepts to API responses. Similarly, static configuration fragments may be generated by templating tools, etc. The APIs and tools used to generate xDS configuration are beyond the scope of the definitions in this repository.

  • REST-JSON API equivalents will be provided for the basic singleton xDS subscription services CDS/EDS/LDS/RDS/SDS. Advanced APIs such as HDS, ADS and EDS multi-dimensional LB will be gRPC only. This avoids having to map complicated bidirectional stream semantics onto REST.

  • Listeners will be immutable. Any updates to a listener via LDS will require the draining of existing connections for the specific bound IP/port. As a result, new requests will only be guaranteed to observe the new configuration after existing connections have drained or the drain timeout.

  • Versioning will be expressed via proto3 package namespaces, i.e. package envoy.api.v2;.

  • Custom components (e.g. filters, resolvers, loggers) will use a reverse DNS naming scheme, e.g. com.google.widget, com.lyft.widget.

APIs

Unless otherwise stated, the APIs with the same names as v1 APIs have a similar role.

In addition to the above APIs, an aggregation API will be provided to allow for fine grained control over the sequencing of API updates across discovery services:

A protocol description for the xDS APIs is provided here.

Terminology

Some relevant existing terminology is repeated below and some new v2 terms introduced.

  • Cluster: A cluster is a group of logically similar endpoints that Envoy connects to. In v2, RDS routes points to clusters, CDS provides cluster configuration and Envoy discovers the cluster members via EDS.

  • Downstream: A downstream host connects to Envoy, sends requests, and receives responses.

  • Endpoint: An endpoint is an upstream host that is a member of one or more clusters. Endpoints are discovered via EDS.

  • Listener: A listener is a named network location (e.g., port, unix domain socket, etc.) that can be connected to by downstream clients. Envoy exposes one or more listeners that downstream hosts connect to.

  • Locality: A location where an Envoy instance or an endpoint runs. This includes region, zone and sub-zone identification.

  • Management server: A logical server implementing the v2 Envoy APIs. This is not necessarily a single physical machine since it may be replicated/sharded and API serving for different xDS APIs may be implemented on different physical machines.

  • Region: Geographic region where a zone is located.

  • Sub-zone: Location within a zone where an Envoy instance or an endpoint runs. This allows for multiple load balancing targets within a zone.

  • Upstream: An upstream host receives connections and requests from Envoy and returns responses.

  • xDS: CDS/EDS/HDS/LDS/RLS/RDS/SDS APIs.

  • Zone: Availability Zone (AZ) in AWS, Zone in GCP.