...

Text file src/sigs.k8s.io/gateway-api/geps/gep-1426.md

Documentation: sigs.k8s.io/gateway-api/geps

     1# GEP-1426: xRoutes Mesh Binding
     2
     3* Issue: [#1294](https://github.com/kubernetes-sigs/gateway-api/issues/1294)
     4* Status: Experimental
     5
     6> **Note**: This GEP is exempt from the [Probationary Period][expprob] rules
     7> of our GEP overview as it existed before those rules did, and so it has been
     8> explicitly grandfathered in.
     9
    10[expprob]:https://gateway-api.sigs.k8s.io/geps/overview/#probationary-period
    11
    12## Overview
    13
    14Similar to how `xRoutes` bind to `Gateways` and manage North/South traffic flows in Gateway API’s ingress use-case, it would be natural to adopt a similar model for traffic routing concerns in service mesh deployments. The purpose of this GEP is to add a mechanism to the Gateway API spec for the purpose of associating the various `xRoute` types to a service mesh and offering a model for service owners to manage traffic splitting configurations.
    15
    16This GEP is intended to establish an implementable, but experimental, baseline for supporting basic service mesh traffic routing functionality through the Gateway API spec.
    17
    18## Personas
    19
    20This GEP uses the [roles and personas](https://gateway-api.sigs.k8s.io/concepts/security-model/#roles-and-personas) defined in the Gateway API security model, and the service "producer" and "consumer" roles defined in [GEP-1324: Service Mesh in Gateway API](https://gateway-api.sigs.k8s.io/geps/gep-1324/#producer-and-consumer).
    21
    22## Goals
    23
    24* MUST allow `xRoute` traffic rules to be configurable for a mesh service by the application owner/producer.
    25* SHOULD allow control by the cluster operator (mesh administrator) to grant permission for whether `xRoute` resources in a given namespace are allowed to configure mesh traffic routing.
    26* SHOULD NOT require downstream "consumer" services to update configuration or DNS addresses for traffic to follow "producer" mesh routing rules configured by upstream services.
    27* SHOULD NOT require reconfiguring existing `xRoute` resources for North/South Gateway configuration.
    28
    29## Non-Goals
    30
    31* Supporting "egress" use cases, which is currently a deferred goal, including:
    32    * Defining how "consumer" traffic rules which could override routing for service upstreams only within the local scope of a namespace or service might be configured.
    33    * Redirecting calls from arbitrary custom domains to an in-cluster service.
    34* Defining how multiple `Services` or `EndpointSlices` representing instances of a single "logical" service should present an identity for AuthN/AuthZ or be associated with each other beyond routing rules.
    35* Defining how AuthZ should be implemented to secure East/West traffic between services.
    36* Defining how [Policy Attachment](https://gateway-api.sigs.k8s.io/reference/policy-attachment/) would bind to `xRoute`, services or a mesh.
    37* Defining how `Routes` configured for East/West service mesh traffic management might integrate with North/South `Gateways`.
    38    * This is a bit tricky in that it's effectively a form of delegation as described in [GEP-1058: Route Inclusion and Delegation](https://github.com/kubernetes-sigs/gateway-api/pull/1085), and is planned to be explored in a future GEP.
    39* Handling East/West traffic outside the cluster (VMs, etc).
    40
    41## Implementation Details and Constraints
    42
    43* MUST set a status field on `xRoute` to show if the routing configuration has been applied to the mesh.
    44* MUST only be allowed to configure "producer" traffic rules for a `Service` in the same namespace as the `xRoute`.
    45    * Traffic routing configuration defined in this way SHOULD be respected by ALL consumer services in all namespaces in the mesh.
    46* MAY assume that a mesh implements "transparent proxy" functionality to redirect calls to the Kubernetes DNS address for a `Service` through mesh routing rules.
    47
    48## Introduction
    49
    50It is proposed that an application owner should configure traffic rules for a mesh service by configuring an `xRoute` with a Kubernetes `Service` resource as a `parentRef`.
    51
    52This approach is dependent on both the "frontend" role of the Kubernetes `Service` resource as defined in [GEP-1324: Service Mesh in Gateway API](https://gateway-api.sigs.k8s.io/geps/gep-1324/#service) when used as a `parentRef` and the "backend" role of `Service` when used as a `backendRef`. The conformant implementation would use the Kubernetes `Service` name to match traffic for meshes, but the `backendRef` endpoints would ultimately be used for the canonical IP address(es) to which traffic should be redirected by rules defined in this `xRoute`. This approach leverages the existing points of extensibility within the Gateway API spec, and would not require introducing any API changes or new resources, only defining expected behavior.
    53
    54### Why Service?
    55
    56The GAMMA initiative has been working to bring service mesh use-cases to the Gateway API spec, taking the best practices and learnings from mesh implementations and codifying them in a spec. Most mesh users are familiar with using the Kubernetes `Service` resource as the foundation for traffic routing. Generally, this architecture makes perfect sense; unfortunately, `Service` is far too coupled of a resource. It orchestrates IP address allocation, DNS, endpoint collection and propagation, load balancing, etc. For this reason, it **cannot** be the right long-term answer for `parentRef` binding; however, it is the only feasible option that Kubernetes has for mesh implementations today. We expect this to change (indeed, we hope to be a part of that change), but in the interest of developing a spec now, we must once again lean on the `Service` resource. However, we will provide provisions to support additional resources as a `parentRef`.
    57
    58## API
    59
    60```yaml
    61metadata:
    62  name: foo-route
    63  namespace: store
    64spec:
    65  parentRefs:
    66  - kind: Service
    67    name: foo
    68  rules:
    69    backendRefs:
    70    - kind: Service
    71      name: foo
    72      weight: 90
    73    - kind: Service
    74      name: foo-v2
    75      weight: 10
    76```
    77
    78In the example above, routing rules have been configured to direct 90% of traffic for the `foo` `Service` to the default "backend" endpoints specified by the `foo` `Service` [`selector`](https://kubernetes.io/docs/concepts/services-networking/service/#defining-a-service) field, and 10% to the `foo-v2` `Service`. This is determined based on the `ClusterIP` (for `Service`) and `ClusterSetIP` (for `ServiceImport`) matching, and for "transparent proxy" mesh implementations would match all requests to `foo.svc.cluster.local` (or arbitrary custom suffix, as the hostname is not specified manually) from within the same namespace, all requests to `foo.store.svc.cluster.local` from other namespaces, and all requests to `foo.store.svc.clusterset.local` for multicluster services, within the scope of the service mesh.
    79
    80### Route presence
    81
    82When no `xRoute` resources are defined, all traffic should implicitly work - this is just how Kubernetes functions. When you create an `xRoute` targeting a service as a `parentRef` you are replacing that implicit logic - not adding to it. Therefore, you may be reshaping or restricting traffic via an `xRoute` configuration (which should be noted, is distinct from *disallowing* traffic by AuthZ).
    83
    84### Allowed service types
    85
    86Services valid to be selected as a `parentRef` SHOULD have a way to identify traffic to them - typically by one or more virtual IP(s), DNS hostname(s), or name(s).
    87
    88Implementations SHOULD support the default [`ClusterIP`](https://kubernetes.io/docs/concepts/services-networking/service/#publishing-services-service-types) `Service` type as a `parentRef`, with or without selectors.
    89
    90["Headless" `Services`](https://kubernetes.io/docs/concepts/services-networking/service/#headless-services) SHOULD NOT be supported as a `parentRef`, because they do not implement the "frontend" functionality of a service.
    91
    92`Service` resource with [`type: NodePort`](https://kubernetes.io/docs/concepts/services-networking/service/#type-nodeport) or [`type: LoadBalancer`](https://kubernetes.io/docs/concepts/services-networking/service/#loadbalancer) MAY be allowed as `parentRefs` or `backendRefs`, as these do provision virtual IPs and are effectively `ClusterIP` services with additional functionality, but it should typically be preferred to expose services publicly through the North/South Gateway API interfaces instead.
    93
    94`Service` resources with [`type: ExternalName`](https://kubernetes.io/docs/concepts/services-networking/service/#externalname) SHOULD NOT be allowed as `parentRefs` or `backendRefs` due to [security concerns](https://github.com/kubernetes/kubernetes/issues/103675), although might eventually play some role in [configuring egress functionality](https://github.com/kubernetes-sigs/gateway-api/issues/1070).
    95
    96Services supported as `backendRefs` SHOULD be consistent with expectations for North/South Gateway API implementations, and MUST have associated endpoints. `ClusterIP` `Services` with selectors SHOULD be supported as a `backendRef`.
    97
    98#### `Service` without selectors
    99
   100An alternate pattern additionally supported by this approach would be to target a `Service` without selectors as the `parentRef`. This could be a clean way to create a pure routing construct and abstract a logical frontend, as traffic would resolve to a `backendRef` `Service` with selectors defined on the `HTTPRoute`, or receive a 4xx/5xx error response if no matching path or valid backend was found.
   101
   102### `parentRef` Conformance Levels
   103
   104Currently (v0.7.0), this spec only considers the `Service` resource to be under Core conformance as a `parentRef`. However, Service is not the only resource that can fulfill the frontend role. While the Gateway API spec couldn’t possibly enumerate every existing (and future) frontend-like resource, it can specify a subset of resources that implementations MUST support as parentRefs under as a part of core conformance. Meshes MAY support other implementation-specific resources as parentRefs. The spec maintainers also reserve the right to add additional resources to core conformance as the spec evolves.
   105
   106#### Extended Conformance
   107
   108In addition to Service, there are other optional parentRef resources that, if used by implementations, MUST adhere to the spec’s prescriptions. At the time of writing (v0.7.0), there is one resource in extended conformance: `ServiceImport` (part of the [MCS API](https://github.com/kubernetes-sigs/mcs-api), currently in alpha). The semantics of `ServiceImport` `parentRef` binding can be found in [GEP-1748](https://gateway-api.sigs.k8s.io/geps/gep-1748/) (Note: Headless `ServiceImport` is out of scope and not currently a part of the spec).
   109
   110##### Why not `IPAddress`
   111
   112In Kubernetes 1.27, there will be a new IPAddress resource added to networking.k8s.io/v1alpha1 as part of [KEP 1880](https://github.com/kubernetes/enhancements/tree/master/keps/sig-network/1880-multiple-service-cidrs#proposal). Naturally, it makes sense to examine whether or not this new resource makes sense as a GAMMA aware parentRef. At first glance, IPAddress seems to be an appropriate abstraction for the “frontend” role we’ve been discussing; every Kubernetes Service is accessed over the network via one of its ip addresses. Furthermore, the fact that the Service resource auto-creates an IPAddress is encouraging. However, the fact that the name of the IPAddress is simply the decimal/hex ip address and not a human-readable Service name makes the UX untenable as a spec-supported parentRef. However, `IPAddress` is NOT disallowed; implementations may use it if they wish.
   113
   114#### Implementation-specific `parentRef`s
   115
   116If mesh implementations wish to enable an implementation-specific resource as a parentRef, they may do so as long as that resource meets certain conditions. Recall that the frontend role of a (generic) service is how one calls the service. In the service mesh transparent proxy context, the frontend role (and parentRef by extension) is effectively the matching mechanism for the specified route. For the Service parentRef, this means that the mesh should apply a particular xRoute’s configuration if the destination ip address for a given connection is the ClusterIP of that parentRef Service. If a mesh wishes to use an implementation-specific resource for parentRef, that resource MUST contain layer-appropriate information suitable for traffic matching (e.g. no Host header capture in TCPRoute). For example, the following HTTPRoute with an Istio `ServiceEntry` as a parentRef would be a valid implementation-specific reference:
   117
   118```yaml
   119apiVersion: networking.istio.io/v1beta1
   120kind: ServiceEntry
   121metadata:
   122  name: internal-svc-httpbin
   123  namespace : egress
   124spec:
   125  hosts:
   126  - example.com
   127  exportTo:
   128  - "."
   129  location: MESH_INTERNAL
   130  ports:
   131  - number: 80
   132    name: http
   133    protocol: HTTP
   134  resolution: DNS
   135---
   136apiVersion: gateway.networking.k8s.io/v1beta1
   137kind: HTTPRoute
   138metadata:
   139  name: mongo-internal
   140spec:
   141  parentRefs:
   142  - kind: ServiceEntry
   143    group: networking.istio.io/v1beta1
   144    name: internal-svc-httpbin
   145    namespace: egress
   146    sectionName: http # referencing the port name
   147  rules:
   148  - backendRefs:
   149    - name: internal-example
   150      port: 80
   151```
   152
   153##### `Gateways`
   154
   155There has been much discussion around cluster local Gateways (i.e. Gateways not associated with a traditional load balancer). While there are various potential UX impairments (e.g. what’s the difference between a GAMMA HTTPRoute with a Gateway parentRef and an ingress implementation’s HTTPRoute?), there is no technical reason why a Gateway cannot be a valid GAMMA parentRef if an implementation wishes to do so.
   156
   157### Route types
   158
   159All types currently defined in the gateway-api core (`HTTP`, `GRPC`, `TCP`, `TLS`, and `UDP`) are available for use in a Mesh implementation.
   160
   161If multiple routes with different types both bind to the same Service and Port pair, only a single route type should be applied. The rejected routes should be ignored and have the `RouteConditionAccepted` status set to the (new) reason `RouteReasonConflicted`.
   162
   163Route type specificity is defined in the following order (first one wins):
   164
   1651. GRPCRoute
   1662. HTTPRoute
   1673. TLSRoute
   1684. TCPRoute
   169
   170Because UDP is its own protocol, it is orthogonal to these precedence order. Since there is only one UDP-based route, there is currently no conflicts possible; if other UDP-based routes are added a similar ordering will be defined.
   171
   172Note: these conflicts only occur when multiple *different* route types apply to the same Service+Port pair. Multiple routes of the same type are valid, and merge according to the route-specific merging semantics.
   173
   174### Ports
   175
   176By default, a `Service` attachment applies to all ports in the service. Users may want to attach routes to only a *specific* port in a Service. To do so, the `parentRef.port` field should be used.
   177
   178If `port` is set, the implementation MUST associate the route only with that port.
   179If `port` is not set, the implementation MUST associate the route with all ports defined in the Service.
   180
   181### `hostnames` field
   182
   183GAMMA implementations SHOULD NOT infer any functionality from the `hostnames` field on `xRoute`s (currently, `TLSRoute`, `HTTPRoute`, and `GRPCRoute` have this field) due to current under-specification and reserved potential for future usage or API changes.
   184
   185For the use case of filtering incoming traffic from selected HTTP hostnames, it is recommended to guide users toward configuring [`HTTPHeaderMatch`](https://gateway-api.sigs.k8s.io/v1alpha2/reference/spec/#gateway.networking.k8s.io%2fv1beta1.HTTPHeaderMatch) rules for the [`Host`](https://developer.mozilla.org/en-US/docs/Web/HTTP/Headers/Host) header. Functionality to be explored in future GEPs may include supporting concurrent usage of an `xRoute` traffic configuration for multiple North/South `Gateways` and East/West mesh use cases or redirection of egress traffic to an in-cluster `Service`.
   186
   187### Namespace boundaries
   188
   189In a mesh, routes can be configured by two personas:
   190
   191* Service producers, who want to modify behavior of *inbound* requests to their services
   192* Service consumers, who want to modify behavior of *outbound* requests to other services.
   193
   194While these concepts are not directly exposed in the API, a route is implicitly fulfilling one of these roles and behaves differently depending on the role.
   195
   196A route is a producer route when the `parentRef` refers to a service in the *same namespace*. This route SHOULD apply to all incoming requests to the service, including from clients in other namespaces.
   197
   198Note: Some implementations may only be able to apply routes on client-side proxies. As a result, these will likely only apply to requests from clients who are also in the mesh.
   199
   200A route is a consumer route when the `parentRef` refers to a service in *another namespace*. Unlike producer routes, consumer routes are scoped only the same namespace. This ensures that for traffic between two namespaces, another unrelated namespace cannot modify their traffic.
   201
   202Routes of either type can send traffic to `backendRefs` in any namespace. Unlike `Gateway` bound routes, this is allowed without a `ReferenceGrant`. In `Gateway`-bound routes (North-South), routes are opt-in; by default, no Services are exposed (often to the public internet), and a service producer must explicitly opt-in by creating a route themselves, or allowing another namespace to via `ReferenceGrant`. For mesh, routes *augment* existing Services, rather than exposing them to a broader scope. As a result, a `ReferenceGrant` is not required in most mesh implementations. Access control, if desired, is handled by other mechanism such as `NetworkPolicy`. While uncommon, if a mesh implementation *does* expose the ability to access a broader scope than would otherwise be reachable, then `ReferenceGrant` must be used for cross namespaces references.
   203
   204### Multiple routes for a Service
   205
   206A service may be used as a `parentRef` (where we attach to the "Service Frontend") or as a `backendRef` (where we attach to the "Service Backend").
   207
   208In general, when a request is sent to a Service frontend (ex: `curl svc`), it should utilize a Route bound to that Service.
   209However, when sent to a Service backend (ex: `curl pod-ip`), it would not.
   210
   211Similarly, if we have multiple "levels" of Routes defined, only the first will be used, as that is the only one that accesses the Service frontend.
   212
   213Consider a cluster with routes for a Service in both a Gateway, consumer namespace, and producer namespace:
   214
   215* Requests from the Gateway will utilize the (possibly merged) set of routes attached to the Gateway
   216* Requests from a namespace with consumer routes will utilize the (possibly merged) set of routes in the consumer namespace
   217* Requests from other namespaces will utilize the (possibly merged) set of routes in the producer namespace
   218
   219The merging of routes occurs only within groups of the same type of routes (Gateway bound, producer, or consumer), and follows the standard route merging behavior already defined.
   220
   221Note: a possible future extension is to allow `backendRefs` to explicitly target a "frontend" or "backend". This could allow chaining multiple routes together. However, this is out of scope for the current GEP.
   222
   223### Drawbacks
   224
   225* The fact that this pattern is used for mesh configuration is implicit - this may benefit from some additional configuration to map the `HTTPRoute` to a particular mesh implementation rather than being picked up by any or all GAMMA meshes present in a cluster. Possible approaches include:
   226* [GEP-1282: Describing Backend Properties](https://gateway-api.sigs.k8s.io/geps/gep-1282/) may be one path to associating a `Service` with a mesh, but likely wouldn't be able to handle the application of multiple `HTTPRoutes` for the same `Service`, but each intended for different mesh implementations
   227    * It's currently unclear how relevant this constraint may be, but associating an `HTTPRoute` with a mesh by this method would additionally require an extra graph traversal step.
   228* Expecting a `Mesh` `parentRef` or similar reference as proposed in [GEP-1291: Mesh Representation](https://docs.google.com/document/d/1oyA9uUH7pNNxxwy3WZGSWx-edHDBLrujcezr8q3el70/edit#) may be a preferred eventual path forward, but wouldn't be required initially, with the assumption that only one mesh should typically be present in a cluster.
   229* No mechanism for egress redirection of traffic from arbitrary hostnames to a mesh service within this approach (but could still be implemented separately).
   230
   231## Alternatives
   232
   233### New `MeshService` (or `HttpService`, `VirtualService`, `ServiceBinding`) resource as `parentRef`
   234
   235Introduce a new resource to represent the "frontend" role of a service as defined in [GEP-1291: Mesh Representation](https://docs.google.com/document/d/1oyA9uUH7pNNxxwy3WZGSWx-edHDBLrujcezr8q3el70/edit#).
   236
   237#### Controller manages new DNS hostname
   238
   239A controller could create a matching selector-less `Service` (i.e. no endpoints), to create a `.cluster.local` name, or could interact with [external-dns](https://github.com/kubernetes-sigs/external-dns) to create a DNS name in an owned domain.
   240
   241Ownership/trust would remain based on naming pattern: `serviceName.namespace.svc.[USER_DOMAIN]`
   242
   243Separate `HttpService`, `TlsService` and `TcpService` resources could have the benefit of allowing us to define protocol specific elements to the spec along with an embedded `CommonServiceSpec`, similar to [`CommonRouteSpec`](https://gateway-api.sigs.k8s.io/v1alpha2/reference/spec/#gateway.networking.k8s.io/v1beta1.CommonRouteSpec), and keep similar patterns as `Service`.
   244
   245##### Drawbacks
   246
   247* May require reconfiguring existing applications to point to a new mesh service hostname - adoption wouldn't be "transparent".
   248* The pattern of creating a new pure routing construct would still be implementable following the proposed approach, by manually creating and targeting a new `Service` without selectors as a `parentRef`, without the overhead of introducing a new resource.
   249
   250#### Manage DNS by binding to an existing `Service`
   251
   252A new `ServiceBinding` resource would directly reference an existing `Service` to determine which traffic should be intercepted and redirected following configured service mesh routing rules and facilitate "transparent proxy" functionality. This resource could possibly share similar responsibilities as the need identified in [GEP-1282: Describing Backend Properties](https://gateway-api.sigs.k8s.io/geps/gep-1282/).
   253
   254```
   255kind: ServiceBinding
   256metadata:
   257  name: foo_binding
   258spec:
   259  parentRefs:
   260  - kind: Service
   261    name: foo
   262---
   263spec:
   264  parentRefs:
   265  - kind: ServiceBinding
   266    name: foo_binding
   267  rules:
   268    backendRefs:
   269    - kind: Service
   270      name: foo
   271      weight: 90
   272    - kind: Service
   273      name: foo_v2
   274      weight: 10
   275```
   276
   277While the `HTTPRoute` does not directly reference a particular mesh implementation in this approach, it would be possible to design the `ServiceBinding` resource to specify that.
   278
   279##### Drawbacks
   280
   281* Introduces an extra layer of abstraction while still having several of the same fundamental drawbacks as a direct `parentRef` binding to `Service`.
   282* May require reconfiguring `Gateway` `HTTPRoutes` to specify `ServiceBindings` as `backendRefs`.
   283
   284#### Drawbacks
   285
   286* The split frontend/backend role of `Service` is fundamentally an issue with the `Service` resource, and while upstream changes may be quite slow, this would likely be best addressed through an upstream KEP - introducing a new resource to GAMMA now would likely result in API churn if we expect a similar proposal to be upstreamed eventually.
   287* Adopting the proposed `Service` as `parentRef` approach wouldn't foreclose the possibility of migrating to a new frontend-only resource in the future, and wouldn't even require a breaking change to `HTTPRoute`, just adding support for a new `parentRef` `Kind`.
   288* Would be less clear how to integrate with transparent proxy functionality - it may be possible to design some way to select a `Service` or hostname to intercept, but abstraction through a separate resource would make configuration more complex.
   289
   290### `Mesh` resource as `parentRef`
   291
   292This binds an `HTTPRoute` directly to a cluster-scoped `Mesh` object as defined in [GEP-1291: Mesh Representation](https://docs.google.com/document/d/1oyA9uUH7pNNxxwy3WZGSWx-edHDBLrujcezr8q3el70/edit#).
   293
   294```
   295spec:
   296  parentRefs:
   297  - kind: Mesh
   298    name: cool-mesh
   299```
   300
   301It is currently undefined how this approach may interact with either explicitly configured [`hostnames`](https://gateway-api.sigs.k8s.io/v1alpha2/reference/spec/#gateway.networking.k8s.io/v1beta1.HTTPRouteSpec) or implicit "transparent proxy" routing for Kubernetes `Services` to determine how traffic should be intercepted and redirected.
   302
   303This approach is not entirely abandoned, as it could supplement the proposed approach if explicit attachment to a specific mesh is deemed necessary. Additionally, this approach may offer a future option for attaching an `HTTPRoute` to a mesh, but not a specific service (e.g. to implement mesh-wide egress functionality for all requests to a specific hostname).
   304
   305#### Peer to `Service` resource `parentRef`
   306
   307An `HTTPRoute` could specify a `Mesh` resource `parentRef` as a peer to a `Service` resource `parentRef`.
   308
   309```
   310spec:
   311  parentRefs:
   312  - kind: Mesh
   313    name: cool-mesh
   314  - kind: Service
   315    name: foo
   316```
   317
   318##### Drawbacks
   319
   320* Would require separate `HTTPRoute` resources to explicitly define _different_ traffic routing rules for the same service on different meshes.
   321
   322#### Nested `services` and `hostnames` fields in [`ParentReference`](https://gateway-api.sigs.k8s.io/v1alpha2/reference/spec/#gateway.networking.k8s.io/v1beta1.ParentReference)
   323
   324In core conformance, the `services` would only be valid for `Mesh` types, and `hostnames` field only for `Gateway`. Mesh implementations could still use a `Host` header match if they wanted limit rules to specific hostnames.
   325
   326```
   327parentRefs:
   328- kind: Mesh
   329  name: coolmesh
   330  services:
   331  - name: foo
   332    kind: Service
   333- kind: Gateway
   334  name: staging
   335  hostnames: [staging.example.com]
   336- kind: Gateway
   337  name: prod
   338  hostnames: [prod.example.com]
   339# Top level hostnames field removed
   340```
   341
   342Moving the `hostnames` field from `HTTPRoute` to `ParentReference` might introduce a clean path for concurrently using a route across North/South and mesh use cases,  even without introducing the `services` field or a new `Mesh` resource, and even makes pure North/South implementations more flexible by allowing a hostname-per-`Gateway` scope.
   343
   344##### Drawbacks
   345
   346* Substantial API change, impacting even North/South use cases
   347* Extending this functionality to support mesh-wide egress or arbitrary redirection may still require some sort of bidirectional handshake with a `Hostname` resource to support configuration across namespaces and limit conflicting configuration.
   348
   349
   350#### Phantom `parentRef`
   351```
   352spec:
   353  parentRefs:
   354  - kind: Mesh
   355    name: istio
   356```
   357
   358This is done by configuring the `parentRef`, to point to the `istio` `Mesh`. This resource does not actually exist in the cluster and is only used to signal that the Istio mesh should be used. In Istio's [experimental implementation](https://istio.io/latest/docs/tasks/traffic-management/ingress/gateway-api/#mesh-traffic), the `hostnames` field on `HTTPRoute` is used to match mesh service traffic to the routing rules.
   359
   360### New field on `HTTPRoute` for `Service` binding
   361
   362A new field `serviceBinding` would be added to `HTTPRoute` to attach to the `Service`. Alternatively, this could be a new field in [`HTTPRouteMatch`](https://gateway-api.sigs.k8s.io/v1alpha2/reference/spec/#gateway.networking.k8s.io/v1beta1.HTTPRouteMatch). As with the proposed implementation, this approach could be combined with a `Mesh` resource or similar as the `parentRef`, which would just define that the route would be applied to a mesh.
   363
   364```
   365spec:
   366  serviceBinding:
   367    name: my-service
   368```
   369
   370OR
   371
   372```
   373spec:
   374  matches:
   375    service:
   376      name: my-service
   377```
   378
   379For either implementation, the type of the `serviceBinding` or `service` field should likely be a struct with `Group` (defaulting to the Kubernetes core API group when unspecified), `Kind` (defaulting to `Service` when unspecified) and `Name` fields, to allow for extensibility to `ServiceImport` or custom mesh service types.
   380
   381#### Drawbacks
   382
   383* API addition required, which is somewhat awkwardly ignored for North/South use cases, and could complicate potential for concurrent use of an `HTTPRoute` across both North/South and mesh use cases.
   384* Adding new fields to a relatively stable resource like `HTTPRoute` could be difficult to do in an experimental way.
   385* Following this pattern may lead to subsequent additional fields to further clarify or extend behavior.
   386
   387### `Gateway` resource with `class: mesh` as `parentRef`
   388
   389To support arbitrary DNS names (owned by a "domain owner persona") we would need a similar mechanism to what `Gateway` is using for delegating management of `HTTPRoutes` to namespaces. Instead of duplicating everything - we could use `Gateway` as is, with `class: mesh` (or matching the mesh implementation desired name).
   390
   391```
   392kind: Gateway
   393spec:
   394  class: mesh
   395  listeners:
   396  - name: example
   397    hostname: "example.com"
   398---
   399kind: HTTPRoute
   400spec:
   401  parentRefs:
   402    name: foo_gateway
   403    sectionName: example
   404  hostnames: ["example.com", "foo.svc.cluster.local"]
   405```
   406
   407Functionally such a mesh could be implemented using the existing gateway spec - a GAMMA implementation would only remove the extra hop through the `Gateway`, using sidecars, or it may use a specialized per-namespace gateway to isolate the mesh traffic (like [Istio Ambient](https://istio.io/latest/blog/2022/introducing-ambient-mesh/)). Proxyless gRPC could also use this to route directly.
   408
   409This solution could work well for both non-`cluster.local` names but also for egress, where a `Gateway` with `class: egress` could define names that are external to the mesh and need to either have policies applied or go to a dedicated egress gateway.
   410
   411#### Drawbacks
   412
   413* Using the `HTTPRoute` `hostnames` field to match mesh traffic breaks from the typical Gateway API pattern of explicit Kubernetes resource references, is extremely implicit, and could reduce portability of configuration.
   414* Potentially unclear translation between conceptual resource and concrete implementation, particularly for "proxyless" mesh implementations.
   415* Service meshes may wish to express [egress](https://istio.io/latest/docs/tasks/traffic-management/egress/egress-gateway/) or [other "in-mesh" gateways](https://www.consul.io/docs/connect/gateways) through an API like this, and it could be confusing to overload this resource too much or conflate different personas who may wish to manage mesh service traffic routing as an application owner separately from egress rules as a service consumer or cluster operator.
   416
   417### `ServiceProjection` resource as `parentRef` and `backendRefs`
   418
   419This approach is similar to the above `ServiceBinding` proposal with a couple of major differences:
   420
   421* `ServiceProjection` encapsulates both "frontend" and "backend" roles of the `Service` resource
   422* `ServiceProjection` could handle the full responsibilities described in [GEP-1282: Describing Backend Properties](https://gateway-api.sigs.k8s.io/geps/gep-1282/)
   423
   424```
   425kind: ServiceProjection
   426metadata:
   427    name: foo
   428    namespace: store
   429spec:
   430    serviceRef:
   431        name: foo
   432        kind: Service|ServiceImport
   433    roles:
   434        frontend:
   435       backend:
   436            loadbalancerConfig:
   437                strategy: RoundRobin
   438             clientTLS:
   439                secretRef:
   440                    ...
   441---
   442kind: HTTPRoute
   443metadata:
   444  name: foo_route
   445  namespace: store
   446spec:
   447  parentRefs:
   448  - kind: ServiceProjection
   449    name: foo
   450    role: frontend
   451  rules:
   452    backendRefs:
   453    - kind: ServiceProjection
   454      name: foo
   455      role: backend
   456      weight: 90
   457    - kind: ServiceProjection
   458      role: backend
   459      name: foo_v2
   460      weight: 10
   461```
   462
   463For convenience, `ServiceProjection` could have a `meshRef` field that, when set instead of `serviceRef`, makes all configuration within the `ServiceProjection` apply to all services in the mesh (the mesh control plane would need to read the `Mesh` resource). Pursuant to the changes to status semantics in [GEP-1364: Status and Conditions Update](https://gateway-api.sigs.k8s.io/geps/gep-1364/), it is necessary for the route to attach to something; in this case, the route attaches to the specific role or profile of the `ServiceProjection` and the mesh control plane should update the route status to reflect that.
   464
   465#### Drawbacks
   466
   467* May require reconfiguring `Gateway` `HTTPRoutes` to specify `ServiceProjections` as `backendRefs`.
   468* Verbose boilerplate for each service.
   469
   470### Implicit backendRef
   471
   472An initial iteration of this GEP had the ability to omit a `backendRef` and have it implicitly be set the same as the `parentRef`.
   473This has been removed due to inconsistency with Gateway `parentRefs` and tight coupling of the "frontend" and "backend" roles.
   474
   475Implementations MUST respect the standard `backendRef` rules as defined by the existing spec.

View as plain text