Docker Cookbook

Enhanced capabilities
A Kubernetes cluster coordinates Docker across multiple nodes and provides a uni‐
fied programming model with enhanced capabilities:
Reliable container restart
Kubernetes can monitor the health of a container and restart it when it fails.
129

---

Self-healing
If a node fails, the Kubernetes management system can automatically reschedule
work onto healthy nodes. Dynamic service membership ensures that those new
containers can be found and used.
High-cluster utilization
By scheduling a diverse set of workloads on a common set of machines, users can
drive dramatically higher utilization compared to static manual placement. The
larger the cluster and the more diverse the workloads, the better the utilization.
Organization and grouping
With large clusters, keeping track of all the containers that are running can be
difficult. Kubernetes provides a flexible labeling system that allows both users
and other systems to think in sets of containers. In addition, Kubernetes supports
the idea of namespaces so different users or teams can have isolated views into
the cluster.
Horizontal scaling and replication
Kubernetes is built to enable easy horizontal scaling. Scaling and load balancing
are intrinsic concepts.
Microservice friendly
Kubernetes clusters are a perfect companion for teams adopting a microservices
architecture. Applications can be broken into smaller parts that are easier to
develop, scale, and reason about. Kubernetes provides ways for a service to find
(commonly called 
discovery
) and communicate with other services.
Streamlined operations
Kubernetes allows for specialized ops teams. Management of the Kubernetes sys‐
tem and the nodes that it runs on can be driven by a dedicated team or out‐
sourced to a cloud service. Operational teams for specific apps (or the develop‐
ment team itself) can then focus on running that application without managing
the details of individual nodes.
New concepts
While Docker is great for dealing with containers running on a single node, Kuber‐
netes addresses the additional challenges around cross-node communication and
scale. To help, Kubernetes introduces a set of new concepts:
Cluster scheduling
The process of placing a container on a specific node to optimize the reliability
and utilization of the cluster.