Transparent Proxying

In order to interecept traffic from and to a service through a kuma-dp data plane proxy instance, Kuma utilizes a variety of patterns.

  • On Kubernetes kuma-dp leverages transparent proxying automatically via iptables or CNI, for all incoming and outgoing traffic that is automatically intercepted by kuma-dp without having to change the application code.
  • On Universal kuma-dp leverages the data plane proxy specification associated to it for receiving incoming requests on a pre-defined port, while it can leverage both transparent proxying and explicit outbound entries in the same data plane proxy specification for all outgoing traffic.

There are several advantages for using transparent proxying in universal mode:

  • Simpler Dataplane resource, as the outbound section becomes obsolete and can be skipped.
  • Uiversal service naming using .mesh DNS domain.
  • Better service manageability (security, tracing).

Preparing the Kuma Control plane

The Kuma control plane exposes a DNS service which handles the name resolution in the .mesh DNS zone. By default it listens on port UDP/5653. For this setup we need it to listen on port UDP/53, therefore make sure that this environment variable is set before running kuma-cp: KUMA_DNS_SERVER_PORT=53.

The IP address of the host that runs Kuma Control plane will be used in the next section. Make sure to have it once, kuma-cp is started.

Setting up the service host

The host that will run the kuma-dp process in transparent proxying mode needs to be prepared with the following steps:

  1. Create a new dedicated user on the machine.
  2. Redirect all the relevant inbound and outbound traffic to the Kuma data plane proxy.
  3. Point the DNS resolving for .mesh to the kuma-cp embedded DNS server.

Kuma comes with kumactl executable which can help us preparing the host. Due to the wide variety of Linux setup options, these steps may vary and may need to be adjusted for the specifics of the particular deployment. However, the common steps would be to execute the following commands as root:

  1. # create a dedicated user called kuma-dp-user
  2. $ useradd -U kuma-dp
  4. # use kumactl
  5. $ kumactl install transparent-proxy \
  6.           --kuma-dp-user kuma-dp \
  7.           --kuma-cp-ip <kuma-cp IP>

Where kuma-dp-user is the name of the dedicated user that will be used to run the kuma-dp process and <kuma-cp IP> is the IP address of the Kuma control plane (kuma-cp).

Please note that this command will change both the host iptables rules as well as modify /etc/resolv.conf, while keeping a backup copy of the original file.

The command has several other options which allow to change the default inbound and outbound redirect ports, add ports for exclusion and also disable the iptables or resolve.conf modification steps. The command’s help has enumerated and documented the available options.

The default settings will exclude the SSH port 22 from the redirection, thus allowing the remote access to the host to be preserved. If the host is set up to use other remote management mechanisms, use --exclude-inbound-ports to provide a comma separated list of the TCP ports that will be excluded from the redirection.

The changes will persist over restarts, so this command is needed only once. Reverting back to the original state of the host can be done by issuing kumactl uninstall transparent-proxy.

Data plane proxy resource

In transparent proxying mode, the Dataplane resource that will be should ommit the networking.outbound section and use networking.transparentProxyingsection instead.

  1. type: Dataplane
  2. mesh: default
  3. name: {{ name }}
  4. networking:
  5.   address: {{ address }}
  6.   inbound:
  7.   - port: {{ port }}
  8.     tags:
  9.       kuma.io/service: demo-client
  10.   transparentProxying:
  11.     redirectPortInbound: 15006
  12.     redirectPortOutbound: 15001

The ports illustrated above are the default ones that kumactl install transparent-proxy will set. These can be changed using the relevant flags to that command.

Invoking the Kuma data plane

It is important that the kuma-dp process runs with the same system user that was passed to kumactl install transparent-proxy --kuma-dp-user.

When systemd is used, this can be done with an entry User=kuma-dp in the [Service] section of the service file.

When starting kuma-dp with a script or some other automation instead, we can use runuser with the aforementioned yaml resource as follows:

  1. $ runuser -u kuma-dp -- \
  2.   /usr/bin/kuma-dp run \
  3.     --cp-address=https://172.19.0.2:5678 \
  4.     --dataplane-token-file=/kuma/token-demo \
  5.     --dataplane-file=/kuma/dpyaml-demo \
  6.     --dataplane-var name=dp-demo \
  7.     --dataplane-var address=172.19.0.4 \
  8.     --dataplane-var port=80  \
  9.     --binary-path /usr/local/bin/envoy

Transparent Proxying

There are two ways of how the service can interact with its sidecar to connect to other services.

One is explicitly defining outbounds in the Dataplane:

  1. type: Dataplane
  2. ...
  3. networking:
  4.   ...
  5.   outbound:
  6.   - port: 10000
  7.     tags:
  8.       kuma.io/service: backend

This approach is simple, but it has the disadvantage that you need to reconfigure the service to use http://localhost:10000 when it wants to connect with service backend. This strategy is used on Universal deployments.

The alternative approach is Transparent Proxying. With Transparent Proxying before we start a service, we apply iptablesTransparent Proxying - 图1 (opens new window) that intercept all the traffic on VM/Pod and redirect it to Envoy. The main advantage of this mode is when you integrate with the current hostname resolving mechanism, you can deploy Service Mesh transparently on the platform without reconfiguring applications.

Kuma provides support for transparent proxying on both Universal and Kubernetes.

Configure intercepted traffic

Kuma deploys iptables rules either with kuma-init Init Container or with cni when deployed with CNI mode.

By default, all the traffic is intercepted by Envoy. You can exclude which ports are intercepted by Envoy with the following annotations placed on the Pod

  1. apiVersion: apps/v1
  2. kind: Deployment
  3. metadata:
  4.   name: example-app
  5.   namespace: kuma-example
  6. spec:
  7.   ...
  8.   template:
  9.     metadata:
  10.       ...
  11.       annotations:
  12.         # all incomming connections on ports 1234 won't be intercepted by Envoy
  13.         traffic.kuma.io/exclude-inbound-ports: "1234"
  14.         # all outgoing connections on ports 5678, 8900 won't be intercepted by Envoy
  15.         traffic.kuma.io/exclude-outbound-ports: "5678,8900"
  16.     spec:
  17.       containers:
  18.         ...

You can also control this value on whole Kuma deployment with the following Kuma CP configuration

  1. KUMA_RUNTIME_KUBERNETES_SIDECAR_TRAFFIC_EXCLUDE_INBOUND_PORTS=1234
  2. KUMA_RUNTIME_KUBERNETES_SIDECAR_TRAFFIC_EXCLUDE_OUTBOUND_PORTS=5678,8900

Global settings can be always overridden with annotations on the individual Pods.

When deploying Kuma with kumactl install control-plane you can set those settings with

  1. kumactl install control-plane \
  2.   --env-var KUMA_RUNTIME_KUBERNETES_SIDECAR_TRAFFIC_EXCLUDE_INBOUND_PORTS=1234
  3.   --env-var KUMA_RUNTIME_KUBERNETES_SIDECAR_TRAFFIC_EXCLUDE_OUTBOUND_PORTS=5678,8900

When deploying Kuma with HELM, use controlPlane.envVar value

  1. envVar:
  2.   KUMA_RUNTIME_KUBERNETES_SIDECAR_TRAFFIC_EXCLUDE_INBOUND_PORTS: "1234"
  3.   KUMA_RUNTIME_KUBERNETES_SIDECAR_TRAFFIC_EXCLUDE_OUTBOUND_PORTS=5678,8900