Learning Path: Scale applications - Production considerations

In this set of tutorials, some of the planning considerations are simplified so that you can focus on learning key Google Kubernetes Engine (GKE) Enterprise edition features and services. Before you start to create your own Google Kubernetes Engine (GKE) Enterprise edition environment similar to the one described in this set of tutorials, there are some additional planning considerations to keep in mind. These considerations include the cluster management level, networking, and availability types.


GKE clusters require careful IP address planning. The networking options that you choose impact the architecture of your GKE clusters. Some of these options can't be changed after they're configured without recreating the cluster.

In this set of tutorials, you use Autopilot mode clusters that always use VPC-native mode networking. VPC-native clusters use alias IP address ranges on GKE nodes, and are required for creating clusters on Shared VPCs. VPC-native clusters scale more easily than routes-based clusters without consuming Google Cloud routes and so are less susceptible to hitting routing limits.

Before you create your own GKE Enterprise environment and deploy workloads, review the following networking guidance:

Cluster modes

In this set of tutorials you create a regional GKE Enterprise cluster that uses Autopilot mode. Autopilot clusters are pre-configured with an optimized cluster configuration that is ready for production workloads. You can alternatively use Standard mode clusters for more advanced configuration flexibility over the underlying infrastructure.

For a more comprehensive overview, review the planning documents that start with the cluster configuration choices.


Namespaces let you organize your applications and isolate components from each other. Each namespace has its own set of resources, such as Pods, Services, and Deployments. For example, you can create a namespace for all your frontend services and a namespace for your backend services. This grouping makes it easier to manage your services and to control access to them.

In this set of tutorials, you deploy the Pods and Services for the Cymbal Bank sample application into a single namespace. This approach reduces deployment complexity, but doesn't let you use namespaces to assign resources to different teams and users, as you might do in a production environment. It also doesn't allow you to take advantage of GKE Enterprise team management features, which help you structure your fleets, namespaces, and permissions to let teams act independently as separate "tenants" on your fleet with their own resources. For a more secure and production-ready example of the Cymbal Bank sample application that uses multiple namespaces, see Cymbal Bank application architecture.

Pod disruption budgets

Pod Disruption Budget (PDB) policies help ensure app performance by preventing Pods going down at the same time when you make a change to the system and limit the number of simultaneously unavailable Pods in a replicated application.

In this set of tutorials, you don't configure and use PDBs. When you complete the tutorial to simulate failure, your Services and nodes should all respond as expected. When you deploy your own workloads, PDBs on nodes might block node draining.

If you use PDBs, review your configuration before you try to cordon and drain nodes. If the nodes can't successfully drain, you might have problems with scheduled maintenance events.

What's next

Get started by completing the first tutorial to a deploy a single GKE cluster that runs a microservices-based application.