Docsright arrowTelepresence OSSright arrowConfiguring intercept using CLI

8 min • read

Configuring intercept using CLI

Specifying a namespace for an intercept

The namespace of the intercepted workload is specified during connect using the --namespace option.

Importing environment variables

Telepresence can import the environment variables from the pod that is being intercepted, see this doc for more details.

Creating an intercept

The following command will intercept all traffic bound to the service and proxy it to your laptop. This includes traffic coming through your ingress controller, so use this option carefully as to not disrupt production environments.

Run telepresence status to see the list of active intercepts.

Finally, run telepresence leave <name of intercept> to stop the intercept.

When intercepting a service that has multiple ports, the name of the service port that has been intercepted is also listed.

If you want to change which port has been intercepted, you can create a new intercept the same way you did above and it will change which service port is being intercepted.

Creating an intercept When multiple services match your workload

Oftentimes, there's a 1-to-1 relationship between a service and a workload, so telepresence is able to auto-detect which service it should intercept based on the workload you are trying to intercept. But if you use something like Argo, there may be two services (that use the same labels) to manage traffic between a canary and a stable service.

Fortunately, if you know which service you want to use when intercepting a workload, you can use the --service flag. So in the aforementioned example, if you wanted to use the echo-stable service when intercepting your workload, your command would look like this:

Intercepting multiple ports

It is possible to intercept more than one service and/or service port that are using the same workload. You do this by creating more than one intercept that identify the same workload using the --workload flag.

Let's assume that we have a service multi-echo with the two ports http and grpc. They are both targeting the same multi-echo deployment.

Port-forwarding an intercepted container's sidecars

Sidecars are containers that sit in the same pod as an application container; they usually provide auxiliary functionality to an application, and can usually be reached at localhost:${SIDECAR_PORT}. For example, a common use case for a sidecar is to proxy requests to a database, your application would connect to localhost:${SIDECAR_PORT}, and the sidecar would then connect to the database, perhaps augmenting the connection with TLS or authentication.

When intercepting a container that uses sidecars, you might want those sidecars' ports to be available to your local application at localhost:${SIDECAR_PORT}, exactly as they would be if running in-cluster. Telepresence's --to-pod ${PORT} flag implements this behavior, adding port-forwards for the port given.

If there are multiple ports that you need forwarded, simply repeat the flag (--to-pod=<sidecarPort0> --to-pod=<sidecarPort1>).

Intercepting headless services

Kubernetes supports creating services without a ClusterIP, which, when they have a pod selector, serve to provide a DNS record that will directly point to the service's backing pods. Telepresence supports intercepting these headless services as it would a regular service with a ClusterIP. So, for example, if you have the following service:

You can intercept it like any other:

Specifying the intercept traffic target

By default, it's assumed that your local app is reachable on 127.0.0.1, and intercepted traffic will be sent to that IP at the port given by --port. If you wish to change this behavior and send traffic to a different IP address, you can use the --address parameter to telepresence intercept. Say your machine is configured to respond to HTTP requests for an intercept on 172.16.0.19:8080. You would run this as:

Replacing a running workload

By default, your application keeps running as Telepresence intercepts it, even if it doesn't receive any traffic (or receives only a subset, as with personal intercepts). This can pose a problem for applications that are active even when they're not receiving requests. For instance, if your application consumes from a message queue as soon as it starts up, intercepting it won't stop the pod from consuming from the queue.

To work around this issue, telepresence intercept allows you to pass in a --replace flag that will stop every application container from running on your pod. When you pass in --replace, Telepresence will restart your application with a dummy application container that sleeps infinitely, and instead just place a traffic agent to redirect traffic to your local machine. The application container will be restored as soon as you leave the intercept.