Using the Docker image on Kubernetes

We provide two examples Kubernetes deployments for the TS ingress controller:

  • example-daemonset.yaml uses a DaemonSet, with an example of using node taints and affinity to run the controller only on master nodes, and exposes TS using a hostPort;
  • example-deployment.yaml uses a Deployment, with a nodePort Service used to expose TS.

These both use the Docker image docker.io/torchbox/k8s-ts-ingress:v1.0.0-alpha8.

You will probably want to read and edit one of these files before using it.

Unfortunately, there are many different ways to expose an Ingress controller on Kubernetes, and we can't document every possible variation, so you will need to decide what method is best for your cluster.

Runtime configuration

Most Traffic Server configuration (records.config entries) can be changed using environment variables; see the Traffic Server Documentation. For example, to make Traffic Server listen on port 7080 for HTTP requests, set PROXY_CONFIG_HTTP_SERVER_PORTS=7080.

The TS Docker image provides one additional environment variable:

  • TS_CACHE_SIZE=nnn: Size of the on-disk cache file to create, in megabytes.

Cache storage

The example deployment resources use an emptyDir for cache storage. This means the cache will persist across node reboots, but will we cleared if the pod is move to a different node, or if it's upgraded (which deletes the old pod).

The cache file will be created automatically on startup if it doesn't exist.

For persistent cache storage, mount a volume on /var/lib/trafficserver. However, be aware that only one instance of TS can access the cache at once. If you are running multiple copies, you will need to create a separate PV for each instance (perhaps by using a StatefulSet instead of a DaemonSet).