Google has announced the general availability of the 1.10 version of their Kubernetes Engine. Furthermore, in parallel of the 1.10 release, Google will release several new features to support enterprise use cases, such as the Shared Virtual Private Cloud (VPC), Regional Persistent Disks and Regional Clusters, Node Auto-Repair, and the Horizontal Pod Autoscaler. These features will provide more availability, uptime, and scalability for the Google Kubernetes Engine (GKE).
With Shared VPC, enterprises can better control their network resources with a more fine-grained model. For instance, administrators can delegate specific responsibilities to project administrators like managing and creating kubernetes engine instances and clusters, while keeping centralized control over network resources such as subnets, and firewalls.
Google also improves availability and uptime for the Kubernetes Engine 1.10 through Regional Persistence Disks (PD) and Regional Clusters. With Regional PDs enterprises do not have to worry about application-level replication. These disks provide durable network-attached block storage with synchronous replication of data between two zones in a region. Furthermore, the Regional Clusters will allow enterprises to further enhance availability by creating a Kubernetes Engine cluster with a multi-master, highly-available control plane that spreads their masters across three zones in a region.
The node auto-repair feature, which is now generally available, enables enterprises to improve uptime. The node auto-repair monitors the health of nodes and repairs them when unhealthy. Note that both Regional Persistence Disks (PD) and Regional Clusters will become available soon, according to Google.
Enterprises can now also scale horizontally with the Kubernetes Engine 1.10 through the Horizontal Pod Autoscaler that, in addition to the default CPU usage metric, now supports three custom metric types:
- External, for example, allowing scaling to be based upon external application-specific metrics or data such as an associated Cloud Pub/Sub queue length (a popular requested feature)
- Pods, enabling scaling to be based upon the comparison of a computed average to a target value. For example, scaling can be based on the average number of open connections per Pod
- Object, which allows a single Pod to expose metrics to trigger scaling. For example, a Pod that is running Prometheus can report metrics based on an associated instance of Kafka running in the cluster, which can then be used to scale Pods running the Kafka consumer services
The 1.10 version has been a part of the continuous evolution of the Google Kubernetes Engine since its first release three years ago. Back in 2015, the Kubernetes Engine was a part of Google’s effort to put out a platform for launching and managing web applications using containers. Since it's availability, the adoption of the engine has grown. Google said in the blog post about the release of 1.10:
Kubernetes Engine, the first production-grade managed Kubernetes service, has been generally available since 2015. Core-hours for the service have ballooned: in 2017 Kubernetes Engine core-hours grew 9X year over year, supporting a wide variety of applications. Stateful workload (e.g., databases and key-value stores) usage has increased since the initial launch in 2016, to over 40 percent of production Kubernetes Engine clusters.
Enterprises can start using Google Kubernetes Engine 1.10 now, and pricing details are available at the Kubernetes Engine pricing page.