User FAQ

Where should I get Kubernetes?

This is covered by our getting-started guide.

We recommend you do not use binaries from http://storage.googleapis.com/kubernetes, as these are neither official releases nor maintained.

Command not found?

Kubernetes expects an argv array for the command. If you put your command and its arguments all in one string, that is interpreted as the command name. Put the arguments into separate array elements instead.

Use of environment variables on the command line

As discussed in Docker's documentation, the container's command is not necessarily executed in a shell, in which case environment variables are not expanded in command-line arguments. In particular, non-default commands are specified via Kubernetes's API using the array form of Docker's Cmd argument, which causes the command to be execed directly. You can cause environment variables to be evaluated by simply making your command a shell, with your application's command line as its argument. For example:

{ "command": [ "sh", "-c", "redis-server /etc/redis/redis.conf --slaveof $SERVICE_HOST $REDIS_MASTER_SERVICE_PORT"] }

How can I ensure that at most one of my pods is scheduled on each node?

The Kubernetes scheduler spreads pods with the same labels (e.g., spawned by the same ReplicationController) and pods targeted by the same Service by default. However, if you need a hard guarantee that no two of the pods will schedule on the same node, request a hostPort for one of the containers in the pod. The scheduler will ensure that at most one pod uses a given host port on any node. For example:

apiVersion: v1
kind: Pod
metadata:
  name: hello
  labels:
    app: hello
    environment: testing
spec:
  containers:
   - name: hello
     image: quay.io/kelseyhightower/hello
     ports:
      - containerPort: 80
        hostPort: 80

Why can't my containers find the service they need to connect to?

You need to create the service before creating the pods that will connect to the service.

Also ensure you are using the correct environment variable names.

If neither of these is your problem, see the Debugging-FAQ.

How do I change the size of my cluster?

Change NUM_MINIONS in the platform-specific config-default.sh file, such as GCE's.

Currently, NUM_MINIONS needs to be changed before turning up the cluster.

How do I pull images from Google Container Registry?

Kubernetes has native support for the Google Container Regisry, when running on Google Compute Engine. If you are running your cluster on Google Compute Engine or Google Container Engine, simply use the full image name (e.g. gcr.io/my_project/image:tag) and the kubelet will automatically authenticate and pull down your private image.

How do I pull images from another private registry?

If you are using a private registry that requires authentication, you need to create a .dockercfg file on all of your machines.

Running docker login -u xxx -p yyy -e zzz@zzzz https://registry.com will automatically create dockercfg at $HOME/.dockercfg. You may need to place this in a specific directory on different providers. Further details here

How do I use the Google Cloud Storage backed private registry?

Update February 2015: The Google Container Regisry (see above) is a hosted private Docker registry, backed by Google Cloud Storage.

To use the google/docker-registry container image:

1. Create a replication controller with 1 or more pods and the google cloud storage registry image
2. Create a Kubernetes service with a PortalIP address of your choice within the portal IP range, and use that IP address for the registry name.