DPs and Data Model

When Kuma (kuma-cp) runs, it waits for the data plane proxies to connect and register themselves. In order for a data plane proxy to successfully run, there must exist at least one Mesh in Kuma. By default the system generates a default Mesh when the control-plane is run for the first time.

Dataplane Entity

A Dataplane entity must be passed to kuma-dp when instance attempts to connect to the control-plane. On Kubernetes, this operation is fully automated. On Universal, it must be executed manually.

To understand why the Dataplane entity is required, we must take a step back. As we have explained already, Kuma follow a sidecar proxy model for the data plane proxies, where we have an instance of a data plane proxy for every instance of our services. Each Service and DP will communicate with each other on the same machine, therefore on 127.0.0.1.

For example, if we have 6 replicas of a “Redis” service, then we must have one instances of kuma-dp running alongside each replica of the service, therefore 6 replicas of kuma-dp and 6 Dataplane entities as well.

When we start a new data plane proxy in Kuma, two things have to happen:

1. The data plane proxy needs to advertise what service it is responsible for. This is what the Dataplane entity does.
2. The data plane proxy process needs to start accepting incoming and outgoing requests.

To do this, we have to create a file with a Dataplane definition and pass it to kuma-dp run. This way, data-plane will be registered in the Control Plane and Envoy will start accepting requests.

The registration of the Dataplane includes three main sections that are described below in the Dataplane Specification:

• address IP at which this dataplane will be accessible to other dataplanes
• inbound networking configuration, to configure on what port the DP will listen to accept external requests, specify on what port the service is listening on the same machine (for internal DP <> Service communication), and the Tags that belong to the service.
• outbound networking configuration, to enable the local service to consume other services.

For example, this is how we start a Dataplane for an hypotetical Redis service and then start the kuma-dp process:

cat dp.yaml
type: Dataplane
mesh: default
name: redis-1
networking:
  address: 192.168.0.1
  inbound:
  - port: 9000
    servicePort: 6379
    tags:
      kuma.io/service: redis

kuma-dp run \
  --cp-address=https://127.0.0.1:5678 \
  --dataplane-file=dp.yaml
  --dataplane-token-file=/tmp/kuma-dp-redis-1-token

In the example above, any external client who wants to consume Redis will have to make a request to the DP on address 192.168.0.1 and port 9000, which internally will be redirected to the Redis service listening on address 127.0.0.1 and port 6379.

Now let’s assume that we have another service called “Backend” that internally listens on port 80, and that makes outgoing requests to the redis service:

cat dp.yaml
type: Dataplane
mesh: default
name: 
networking:
  address: 
  inbound:
  - port: 8000
    servicePort: 80
    tags:
      kuma.io/service: backend
      kuma.io/protocol: http
  outbound:
  - port: 10000
    tags:
      kuma.io/service: redis

kuma-dp run \
  --cp-address=https://127.0.0.1:5678 \
  --dataplane-file=dp.yaml \
  --dataplane-var name=`hostname -s` \
  --dataplane-var address=192.168.0.2 \
  --dataplane-token-file=/tmp/kuma-dp-backend-1-token

In order for the backend service to successfully consume redis, we specify an outbound networking section in the Dataplane configuration instructing the DP to listen on a new port 10000 and to proxy any outgoing request on port 10000 to the redis service. For this to work, we must update our application to consume redis on 127.0.0.1:10000.

Envoy

kuma-dp is built on top of Envoy, which has a powerful Admin API that enables monitoring and troubleshooting of a running dataplane.

By default, kuma-dp starts Envoy Admin API on the loopback interface (that is only accessible from the local host) and the first available port from the range 30001-65535.

If you need to override that behaviour, you can use --admin-port command-line option or KUMA_DATAPLANE_ADMIN_PORT environment variable.

E.g.,

• you can change the default port range by using --admin-port=10000-20000
• you can narrow it down to a single port by using --admin-port=9901
• you can turn Envoy Admin API off by using --admin-port=

Tags

Each Kuma data plane proxy is associated with tags - or attributes - that can be used to both identify the service that the data plane proxy is representing, and they can also be used when configuring the service mesh policies to determine their behavior in a more flexible way.

A tag attributes a qualifier to the data plane proxy, and the tags that are reserved to Kuma are prefixed with kuma.io like:

• kuma.io/service: Identifies the service name. On Kubernetes this tag is automatically created, while on Universal it must be specified manually.
• kuma.io/zone: Identifies the zone name in a multi-zone deployment. This tag is automatically created and cannot be overwritten.
• kuma.io/protocol: Identifies the protocol that is being exposed by the service and its data plane proxies. Accepted values are tcp, http, http2, grpc and kafka.

Kubernetes

When Dataplane entity is automatically created, all labels from Pod are converted into Dataplane tags. Labels with keys that contains kuma.io/ are not converted because they are reserved to Kuma. The following tags are added automatically and cannot be overridden using Pod labels.

• kuma.io/service: Identifies the service name based on a Service that selects a Pod. Example: demo-app_kuma-demo_svc_80.
• kuma.io/zone: Identifies the zone name in a multi-zone deployment.
• kuma.io/protocol: Identifies the protocol that was defined on the Service that selects a Pod.
• k8s.kuma.io/namespace: Identifies the Pod’s namespace. Example: kuma-demo.
• k8s.kuma.io/service-name: Identifies the name of Kubernetes Service that selects a Pod. Example: demo-app.
• k8s.kuma.io/service-port: Identifies the port of Kubernetes Service that selects a Pod. Example: 80.

Dataplane Specification

The Dataplane entity includes the networking and naming configuration that a data-plane proxy (kuma-dp) must have attempting to connect to the control-plane (kuma-cp).

In Universal mode we must manually create the Dataplane entity before running kuma-dp. A Dataplane entity can be created with kumactl or the HTTP API. When using kumactl, the regular entity definition will look like:

type: Dataplane
mesh: default
name: web-01
networking:
  address: 127.0.0.1
  inbound:
    - port: 11011
      servicePort: 11012
      tags:
        kuma.io/service: backend
        kuma.io/protocol: http
  outbound:
    - port: 33033
      tags:
        kuma.io/service: redis

And the Gateway mode’s entity definition will look like:

type: Dataplane
mesh: default
name: kong-01
networking:
  address: 10.0.0.1
  gateway:
    tags:
      kuma.io/service: kong
  outbound:
  - port: 33033
    tags:
      kuma.io/service: backend

The Dataplane entity includes a few sections:

• type: must be Dataplane.
• mesh: the Mesh name we want to associate the data-plane with.
• name: this is the name of the data-plane instance, and it must be unique for any given Mesh. We might have multiple instances of a Service, and therefore multiple instances of the sidecar data-plane proxy. Each one of those sidecar proxy instances must have a unique name.
• networking: this is the meaty part of the configuration. It determines the behavior of the data-plane on incoming (inbound) and outgoing (outbound) requests.
  • address IP address or domain name at which this dataplane will be accessible to other dataplanes. Domain name will be resolved to IP in the control plane.
  • inbound: an array of objects that determines what services are being exposed via the data-plane. Each object only supports one port at a time, and you can specify more than one objects in case the service opens up more than one port.
    • port: determines the port at which other dataplanes will consume the service
    • serviceAddress: IP at which the service is listening. Defaults to 127.0.0.1. Typical usecase is Universal mode, where kuma-dp runs ina separate netns, container or host than the service.
    • servicePort: determines the port of the service deployed next to the dataplane. This can be omitted if service is exposed on the same port as the dataplane, but only listening on serviceAddress or 127.0.0.1 and differs from networking.address.
    • address: IP at which inbound listener will be exposed. By default it is inherited from networking.address
    • tags: each data-plane can include any arbitrary number of tags, with the only requirement that kuma.io/service is mandatory and it identifies the name of service. You can include tags like version, cloud, region, and so on to give more attributes to the Dataplane (attributes that can later on be used to apply policies).
  • gateway: determines if the data-plane will operate in Gateway mode. It replaces the inbound object and enables Kuma to integrate with existing API gateways like Kong.
    • tags: each data-plane can include any arbitrary number of tags, with the only requirement that kuma.io/service is mandatory and it identifies the name of service. You can include tags like version, cloud, region, and so on to give more attributes to the Dataplane (attributes that can later on be used to apply policies).
  • outbound: every outgoing request made by the service must also go thorugh the DP. This object specifies ports that the DP will have to listen to when accepting outgoing requests by the service:
    • port: the port that the service needs to consume locally to make a request to the external service
    • address: the IP at which outbound listener is exposed. By default it is 127.0.0.1 since it should only be consumed by the app deployed next to the dataplane.
    • tags: traffic on port:address will be sent to each data-plane that matches those tags. You can put many tags here. However, it is recommended to keep the list short and then use TrafficRoute for dynamic management of the traffic.

Kubernetes

On Kubernetes the data-planes are automatically injected by Kuma as long as the K8s Namespace or Pod are labeled with kuma.io/sidecar-injection = enabled, e.g.

apiVersion: v1
kind: Namespace
metadata:
  name: kuma-example
  labels:
    # inject Kuma sidecar into every Pod in that Namespace,
    # unless a user explicitly opts out on per-Pod basis
    kuma.io/sidecar-injection: enabled

To opt out of data-plane injection into a particular Pod, you need to label it with kuma.io/sidecar-injection = disabled, e.g.

apiVersion: apps/v1
kind: Deployment
metadata:
  name: example-app
  namespace: kuma-example
spec:
  ...
  template:
    metadata:
      ...
      labels:
        # indicate to Kuma that this Pod doesn't need a sidecar
        kuma.io/sidecar-injection: disabled
    spec:
      containers:
        ...

On Kubernetes the Dataplane entity is also automatically created for you, and because transparent proxying is being used to communicate between the service and the sidecar proxy, no code changes are required in your applications.

Gateway

The Dataplane entity can operate in gateway mode. This way you can integrate Kuma with existing API Gateways like Kong.

When you use a data plane proxy with a service, both inbound traffic to a service and outbound traffic from the service flows through the proxy. API Gateway should be deployed as any other service within the mesh. However, in this case we want inbound traffic to go directly to API Gateway, otherwise clients would have to be provided with certificates that are generated dynamically for communication between services within the mesh. Security for an entrance to the mesh should be handled by API Gateway itself.

Gateway mode lets you skip exposing inbound listeners so it won’t be intercepting ingress traffic.

Universal

On Universal, you can define the Dataplane entity like this:

type: Dataplane
mesh: default
name: kong-01
networking:
  address: 10.0.0.1
  gateway:
    tags:
      kuma.io/service: kong
  outbound:
  - port: 33033
    tags:
      kuma.io/service: backend

When configuring your API Gateway to pass traffic to backend set the url to http://localhost:33033

Kubernetes

On Kubernetes, Dataplane entities are automatically generated. To inject gateway Dataplane, mark your API Gateway’s Pod with kuma.io/gateway: enabled annotation. Here is example with Kong for Kubernetes:

apiVersion: apps/v1
kind: Deployment
metadata:
  labels:
    app: ingress-kong
  name: ingress-kong
  namespace: kong
spec:
  template:
    metadata:
      annotations:
        kuma.io/gateway: enabled
    spec:
      containers:
        image: kong:1.3
      ...

The optimal gateway in Kubernetes mode would be Kong. You can use Kong for Kubernetes to implement authentication, transformations, and other functionalities across Kubernetes clusters with zero downtime. Using Kong for Kubernetes with Kuma requires an annotation on every Service that you want to pass traffic to ingress.kubernetes.io/service-upstream=true. This is automatically injected by Kuma for every Kubernetes service that is in a namespace part of the mesh i.e. has kuma.io/sidecar-injection: enabled label.

Services can be exposed to an API Gateway in one specific zone, or in multi-zone. For the latter, we need to expose a dedicated Kubernetes Service object with type ExternalName, which sets the externalName to the .mesh DNS record for the particular service that we want to expose, that will be resolved by Kuma’s internal service discovery.

Example Gateway in Multi-Zone

In this Kubernetes example, we will be exposing a service named frontend-api (that is running on port 8080) and deployed in the kuma-demo namespace. In order to do so, the following Service needs to be created manually:

apiVersion: v1
kind: Service
metadata:
  name: frontend
  namespace: kuma-demo
spec:
  type: ExternalName
  externalName: frontend-api.kuma-demo.svc.8080.mesh

Finally, we need to create the corresponding Kubernetes Ingress resource:

apiVersion: extensions/v1beta1
kind: Ingress
metadata:
  name: frontend
  namespace: kuma-demo
  annotations:
    kubernetes.io/ingress.class: kong
spec:
  rules:
  - http:
      paths:
      - path: /
        backend:
          serviceName: frontend
          servicePort: 80

Note that since we are addressing the service by its domain name frontend-api.kuma-demo.svc.8080.mesh, we should always refer to port 80 (this port is only a placeholder and will be automatically replaced with the actual port of the service).

For an in-depth example on deploying Kuma with Kong for Kubernetes, please follow this demo application guide.

Zone Ingress

To implement cross-zone communication when Kuma is deployed in a multi-zone mode, there is a new proxy type ZoneIngress. These proxies are not attached to any particular workload. Instead, they are bound to that particular zone. All requests that are sent from one zone to another will be directed to the proper instance by the Zone Ingress.

The ZoneIngress entity includes a few sections:

• type: must be ZoneIngress.
• name: this is the name of the Zone Ingress instance, and it must be unique for any given zone.
• networking: contains networking parameters of the Zone Ingress
  • address: the address of the network interface Zone Ingress is listening on. Could be the address of either public or private network interface, but the latter must be used with a load balancer.
  • port: is a port that Zone Ingress is listening on
  • advertisedAddress: an IP address or hostname which will be used to communicate with the Zone Ingress. Zone Ingress doesn’t listen on this address. If Zone Ingress is exposed using a load balancer, then the address of the load balancer should be used here. If Zone Ingress is listening on the public network interface, then the address of the public network interface should be used here.
  • advertisedPort: a port which will be used to communicate with the Zone Ingress. Zone Ingress doesn’t listen on this port.
• availableServices [auto-generated on Kuma CP] : the list of services that could be consumed through the Zone Ingress
• zone [auto-generated on Kuma CP] : zone where Zone Ingress belongs to

Zone Ingress without advertisedAddress and advertisedPort is not taken into account when generating Envoy configuration, because they cannot be accessed by data plane proxies from other zones.

type: ZoneIngress
name: dp-ingress
networking:
  address: 192.168.0.1
  port: 10001
  advertisedAddress: 10.0.0.1
  advertisedPort: 10000

A ZoneIngress deployment can be scaled horizontally. Many instances can have the same advertised address and advertised port because they can be put behind one load balancer.

Direct access to services

By default on Kubernetes data plane proxies communicate with each other by leveraging the ClusterIP address of the Service resources. Also by default, any request made to another service is automatically load balanced client-side by the data plane proxy that originates the request (they are load balanced by the local Envoy proxy sidecar proxy).

There are situations where we may want to bypass the client-side load balancing and directly access services by using their IP address (ie: in the case of Prometheus wanting to scrape metrics from services by their individual IP address).

When an originating service wants to directly consume other services by their IP address, the originating service’s Deployment resource must include the following annotation:

kuma.io/direct-access-services: Service1, Service2, ServiceN

Where the value is a comma separated list of Kuma services that will be consumed directly. For example:

apiVersion: apps/v1
kind: Deployment
metadata:
  name: example-app
  namespace: kuma-example
spec:
  ...
  template:
    metadata:
      ...
      annotations:
        kuma.io/direct-access-services: "backend_example_svc_1234,backend_example_svc_1235"
    spec:
      containers:
        ...

We can also use * to indicate direct access to every service in the Mesh:

kuma.io/direct-access-services: *