open-consul/website/pages/docs/internals/discovery-chain.mdx

240 lines
9.9 KiB
Plaintext
Raw Normal View History

---
2020-04-07 18:55:19 +00:00
layout: docs
page_title: Discovery Chain
sidebar_title: Discovery Chain
2020-04-07 18:55:19 +00:00
description: >-
The service discovery process can be modeled as a "discovery chain" which
passes through three distinct stages: routing, splitting, and resolution. Each
of these stages is controlled by a set of configuration entries.
---
# Discovery Chain
2020-04-07 23:56:08 +00:00
-> **1.6.0+:** This feature is available in Consul versions 1.6.0 and newer.
2020-04-09 23:46:54 +00:00
~> This topic is part of a [low-level API](/api/discovery-chain)
primarily targeted at developers building external [Connect proxy
2020-04-09 23:46:54 +00:00
integrations](/docs/connect/proxies/integrate).
The service discovery process can be modeled as a "discovery chain" which
passes through three distinct stages: routing, splitting, and resolution. Each
of these stages is controlled by a set of [configuration
2020-04-09 23:46:54 +00:00
entries](/docs/agent/config_entries). By configuring different phases of
the discovery chain a user can control how proxy upstreams are ultimately
resolved to specific instances for load balancing.
-> **Note:** The discovery chain is currently only used to discover
2020-04-09 23:20:00 +00:00
[Connect](/docs/connect) proxy upstreams.
## Configuration
The configuration entries used in the discovery chain are designed to be simple
to read and modify for narrowly tailored changes, but at discovery-time the
various configuration entries interact in more complex ways. For example:
2020-04-09 23:46:54 +00:00
- If a [`service-resolver`](/docs/agent/config-entries/service-resolver)
is created with a [service
2020-04-09 23:46:54 +00:00
redirect](/docs/agent/config-entries/service-resolver#service) defined,
then all references made to the original service in any other configuration
entry is replaced with the redirect destination.
2020-04-09 23:46:54 +00:00
- If a [`service-resolver`](/docs/agent/config-entries/service-resolver)
is created with a [default
2020-04-09 23:46:54 +00:00
subset](/docs/agent/config-entries/service-resolver#defaultsubset)
defined then all references made to the original service in any other
configuration entry that did not specify a subset will be replaced with the
default.
2020-04-09 23:46:54 +00:00
- If a [`service-splitter`](/docs/agent/config-entries/service-splitter)
is created with a [service
2020-04-09 23:46:54 +00:00
split](/docs/agent/config-entries/service-splitter#splits), and the target service has its
2020-04-06 20:27:35 +00:00
own `service-splitter` then the overall effect is flattened and only a single
aggregate traffic split is ultimately configured in the proxy.
2020-04-09 23:46:54 +00:00
- [`service-resolver`](/docs/agent/config-entries/service-resolver)
redirect loops must be rejected as invalid.
2020-04-09 23:46:54 +00:00
- [`service-router`](/docs/agent/config-entries/service-router) and
[`service-splitter`](/docs/agent/config-entries/service-splitter)
configuration entries require an L7 compatible protocol be set for the
service via either a
2020-04-09 23:46:54 +00:00
[`service-defaults`](/docs/agent/config-entries/service-defaults) or
[`proxy-defaults`](/docs/agent/config-entries/proxy-defaults) config
entry. Violations must be rejected as invalid.
2020-04-06 20:27:35 +00:00
- If an [upstream
2020-04-09 23:46:54 +00:00
configuration](/docs/connect/registration/service-registration#upstream-configuration-reference)
[`datacenter`](/docs/connect/registration/service-registration#datacenter)
parameter is defined then any configuration entry that does not explicitly
refer to a desired datacenter should use that value from the upstream.
## Compilation
To correctly interpret a collection of configuration entries as a valid
discovery chain, we first compile them into a form more directly usable by the
layers responsible for configuring Connect sidecar proxies.
You can interact with the compiler directly using the [discovery-chain
2020-04-09 23:46:54 +00:00
API](/api/discovery-chain).
### Compilation Parameters
2020-04-06 20:27:35 +00:00
- **Service Name** - The service being discovered by name.
- **Datacenter** - The datacenter to use as the basis of compilation.
- **Overrides** - Discovery-time tweaks to apply when compiling. These should
be derived from either the
2020-04-09 23:46:54 +00:00
[proxy](/docs/connect/registration/service-registration#complete-configuration-example)
or
2020-04-09 23:46:54 +00:00
[upstream](/docs/connect/registration/service-registration#upstream-configuration-reference)
configurations if either are set.
### Compilation Results
The response is a single wrapped `CompiledDiscoveryChain` field:
```json
{
"Chain": {...<CompiledDiscoveryChain>...}
}
```
#### `CompiledDiscoveryChain`
The chain encodes a digraph of [nodes](#discoverygraphnode) and
[targets](#discoverytarget). Nodes are the compiled representation of various
discovery chain stages and targets are instructions on how to use the [health
2020-04-09 23:46:54 +00:00
API](/api/health#list-nodes-for-connect-capable-service) to retrieve
relevant service instance lists.
You should traverse the nodes starting with [`StartNode`](#startnode). The
nodes can be resolved by name using the [`Nodes`](#nodes) field. Targets can be
resolved by name using the [`Targets`](#targets) field.
- `ServiceName` `(string)` - The requested service.
- `Namespace` `(string)` - The requested namespace.
- `Datacenter` `(string)` - The requested datacenter.
- `CustomizationHash` `(string: <optional>)` - A unique hash of any overrides
that affected the compilation of the discovery chain.
2020-04-06 20:27:35 +00:00
If set, this value should be used to prefix/suffix any generated load
balancer data plane objects to avoid sharing customized and non-customized
versions.
- `Protocol` `(string)` - The overall protocol shared by everything in the
chain.
- `StartNode` `(string)` - The first key into the `Nodes` map that should be
followed when traversing the discovery chain.
- `Nodes` `(map<string|DiscoveryGraphNode>)` - All nodes available for traversal in
the chain keyed by a unique name. You can walk this by starting with
`StartNode`.
2020-04-06 20:27:35 +00:00
-> The names should be treated as opaque values and are only guaranteed to be
consistent within a single compilation.
- `Targets` `(map<string|DiscoveryTarget>)` - A list of all targets used in this chain.
2020-04-06 20:27:35 +00:00
-> The names should be treated as opaque values and are only guaranteed to be
consistent within a single compilation.
#### `DiscoveryGraphNode`
A single node in the compiled discovery chain.
- `Type` `(string)` - The type of the node. Valid values are: `router`,
`splitter`, and `resolver`.
- `Name` `(string)` - The unique name of the node.
- `Routes` `(array<DiscoveryRoute>)` - Only set for `Type:router`. List of routes to
render.
- `Definition` `(ServiceRoute)` - Relevant portion of underlying
`service-router`
2020-04-09 23:46:54 +00:00
[route](/docs/agent/config-entries/service-router#routes).
- `NextNode` `(string)` - The name of the next node in the chain in [`Nodes`](#nodes).
- `Splits` `(array<DiscoverySplit>)` - Only set for `Type:splitter`. List of traffic
splits.
- `Weight` `(float32)` - Copy of underlying `service-splitter`
2020-04-09 23:46:54 +00:00
[`weight`](/docs/agent/config-entries/service-splitter#weight) field.
- `NextNode` `(string)` - The name of the next node in the chain in [`Nodes`](#nodes).
- `Resolver` `(DiscoveryResolver: <optional>)` - Only set for `Type:resolver`. How
to resolve the service instances.
- `Default` `(bool)` - Set to true if no `service-resolver` config entry is
defined for this node and the default was synthesized.
- `ConnectTimeout` `(duration)` - Copy of the underlying `service-resolver`
2020-04-09 23:46:54 +00:00
[`ConnectTimeout`](/docs/agent/config-entries/service-resolver#connecttimeout)
field. If one is not defined the default of `5s` is returned.
- `Target` `(string)` - The name of the target to use found in [`Targets`](#targets).
- `Failover` `(DiscoveryFailover: <optional>)` - Compiled form of the
underlying `service-resolver`
2020-04-09 23:46:54 +00:00
[`Failover`](/docs/agent/config-entries/service-resolver#failover)
definition to use for this request.
- `Targets` `(array<string>)` - List of targets found in
[`Targets`](#targets) to failover to in order of preference.
- `LoadBalancer` `(LoadBalancer: <optional>`) - Copy of the underlying `service-resolver`
[`LoadBalancer`](/docs/agent/config-entries/service-resolver#loadbalancer) field.
If a `service-splitter` splits between services with differing `LoadBalancer` configuration
the first hash-based load balancing policy is copied.
#### `DiscoveryTarget`
- `ID` `(string)` - The unique name of this target.
- `Service` `(string)` - The service to query when resolving a list of service instances.
- `ServiceSubset` `(string: <optional>)` - The
2020-04-09 23:46:54 +00:00
[subset](/docs/agent/config-entries/service-resolver#service-subsets) of
the service to resolve.
- `Namespace` `(string)` - The namespace to use when resolving a list of service instances.
- `Datacenter` `(string)` - The datacenter to use when resolving a list of service instances.
- `Subset` `(ServiceResolverSubset)` - Copy of the underlying
`service-resolver`
2020-04-09 23:46:54 +00:00
[`Subsets`](/docs/agent/config-entries/service-resolver#subsets)
definition for this target.
2020-04-06 20:27:35 +00:00
- `Filter` `(string: "")` - The
2020-04-09 23:46:54 +00:00
[filter expression](/api/features/filtering) to be used for selecting
instances of the requested service. If empty all healthy instances are
returned.
- `OnlyPassing` `(bool: false)` - Specifies the behavior of the resolver's
health check interpretation. If this is set to false, instances with checks
in the passing as well as the warning states will be considered healthy. If
this is set to true, only instances with checks in the passing state will
be considered healthy.
- `MeshGateway` `(MeshGatewayConfig)` - The [mesh gateway
configuration](/docs/connect/mesh-gateway#connect-proxy-configuration)
to use when connecting to this target's service instances.
- `Mode` `(string: "")` - One of `none`, `local`, or `remote`.
- `External` `(bool: false)` - True if this target is outside of this consul cluster.
- `SNI` `(string)` - This value should be used as the
[SNI](https://en.wikipedia.org/wiki/Server_Name_Indication) value when
connecting to this set of endpoints over TLS.
- `Name` `(string)` - The unique name for this target for use when generating
load balancer objects. This has a structure similar to [SNI](#sni), but will
not be affected by SNI customizations such as
2020-04-09 23:46:54 +00:00
[`ExternalSNI`](/docs/agent/config-entries/service-defaults#externalsni).