Kubernetes

What is Kubernetes?

Kubernetes is an open source orchestrator that automates the management, placement, scaling, and routing of connections to and among containers. It provides a way of hosting container-based workloads that helps keep applications highly available and performant, despite problems occurring with underlying infrastructure, or with containerized software itself.

Kubernetes manages this, first, by coordinating among many different types of higher-order Kubernetes services, which are called by different names and perform inter-related functions. Collectively, these Kubernetes services work to maintain applications in a desired state, described in aggregate by the configurations of each of the application’s components.

If something happens to change the state — for example, a worker node fails causing many containers to vanish — Kubernetes services will quickly identify the problem, restart the vanished containers on available nodes, and make them accessible to one another, and to other application components, once again. This is very powerful: eliminating on-call crises, and limiting the need for human operators to intervene to keep important applications running in emergencies — all without Kubernetes users needing to write application-specific automation.

As noted, the Kubernetes services that provide core functionality have many names, like API service and scheduling service and load balancing service. Developers can also create custom Kubernetes services — called Operators — to maintain the state of applications and assist in managing them. Most of this functionality, meanwhile, is enabled by another kind of Kubernetes service — one that’s actually called a “Kubernetes service,” among other things.

This is actually a configuration — concretely a YAML or JSON file — that defines aspects of a workload or component, like how many replicas of that workload or component you want to keep running. This kind of Kubernetes service abstracts application configuration away from actual, running application components (containers and pods): you define the services, and let Kubernetes worry about the details of allocating resources and keeping them running.

Besides “Kubernetes services,” Kubernetes services of the latter type are called several different things, depending on what they define and where they exist in Kubernetes hierarchy of abstractions. Specifically, Kubernetes services define a set of Pods (each pod contains and supports one or more running containers) and exposes them for access by other components or external agencies on an IP address or equivalent. Kubernetes services can have one of several different types, including ClusterIP, which makes the service (and its pods) accessible within the cluster, and NodePort, which exposes the service on a desired IP port on all cluster nodes. Another type of Kubernetes services is Loadbalancer, which tells the cluster to negotiate with an external load balancing service (e.g., provided by a host cloud) to provide load balancing for the pods on a fixed IP address.

Mirantis Kubernetes Engine (formerly Docker Enterprise/UCP) provides a secure and fully-conformant Kubernetes distribution for developers and operators of all skill levels. With Mirantis Kubernetes Engine, organizations can run Kubernetes interchangeably with Swarm orchestration for ultimate flexibility at runtime. Deployed clusters can also be attached as children to join fleets of clusters managed by Mirantis Container Cloud, our multi-cloud container platform.