Asterisk on Kubernetes Demo

This repository contains code and markup for the deployment of a highly scalable voice application on Kubernetes using Kamailio, Asterisk, and NATS.

Supporting tools:

• ARI core Go library github.com/CyCoreSystems/ari
• ARI NATS proxy github.com/CyCoreSystems/ari-proxy
• Netdiscover cloud networking discovery tool github.com/CyCoreSystems/netdiscover
• Asterisk Config kubernetes-based Asterisk templating and update engine github.com/CyCoreSystems/asterisk-config
• Kamailio Dispatchers kubernetes-based update tool github.com/CyCoreSystems/dispatchers
• AudioSocket github.com/CyCoreSystems/audiosocket

Getting started

There are a number of kubernetes YAML files in the k8s directory. Some have numerical prefixes indicating that they should be deployed in a particular order. For the most part, getting the demo off the ground is as easy as installing these YAML files using the usual kubectl apply -f <filename.yaml> method. However, there are a few things which must still be done by hand.

Asterisk config

The required configuration for Asterisk has been stripped down a lot, but there are still a few things which need to be set up: ARI, dialplan, and PJSIP. Examples are included in the asteriskconfig directory. However, you will need to update the inbound.conf.tmpl file with your own DIDs (telephone numbers).

Once configured, you will need to load this configuration in to kubernetes.

1. create a .zip file of the contents of the asteriskconfig directory: - cd asteriskconfig - zip -r ../asterisk-config.zip *
2. load that .zip file in as a Secret - kubectl -n voip create secret generic asterisk-config --from-file=asterisk-config.zip

Kamailio nodeSelector

The default kamailio DaemonSet looks for a GKE nodepool named kamailio. If this nodepool does not exist, kamailio will not be scheduled to run anywhere. Therefore, you should either create the nodepool or modify the kamailio DaemonSet to look for a different nodeSelector.

Google Voice API key

If you intend to use the Google Speech APIs demo, you will need your own API key loaded. When you create an API key on Google, you are given the option to download it as a .JSON file. Do so, then load that file in as key.json in a Secret named speech-key.

• kubectl -n voip create secret generic speech-key --from-file=key.json

Firewall rules

Depending on the environment your kubernetes is deployed to, there are any number of ways to configure the firewall. Fundamentally, though, UDP ports 5060 and 10000-30000 need to flow into the nodes on which the kamailio (and rtpproxy) Pods are running.

On GCP, this is fairly easy. You can create a special Node Pool on which the kamailio Pods will be scheduled which have special instance tags applied. Then, you can tell the GCP firewall to allow the UDP ports 5060,10000-30000 into instances with those special tags.

The kamailio deployment currently expects a nodepool to be available and named kamailio in order to schedule kamailio Pods.