- images: each container in a pod has an image. Currently only
dockerimages are supported.
- pod: smallest deployable unit in a kube cluster. Consists of one or more images.
- replica sets (rs): multi-pod hypervisor, makes sure just the right amount of pods are running at any given time.
- deployments: sets of
replication controllers. Can be defined in a declarative format which describes the desired state
- services: persistant wrapper around a
deployment; provides policies, virtual network addresses and labels
- node: single virtual or physical machine in a kube cluster
- cluster: group of nodes firewalled from the internet
- ingress resources: incoming traffic router
replication controllers (rc) used to be on this list, but is
currently in the process of being superseded by
rs. They should be similar in
helm(1) is to kube, what
homebrew is to OS X.
Labels are added onto system objects to provide a multi-tiered structure. The
following labels are enforced by
helm(1), and are probably a good idea to
rely on for non-helm kube structures too:
- group: same role. E.g. frontend, api, data
- provider: type of service provided. E.g. etcd, postgres, s3
- mode: operation mode of the service. E.g. standalone, clustered, discovery