This page explains how regional clusters work. To learn how to create a regional cluster, refer to Creating a Cluster.
By default, a cluster creates its cluster master and its nodes in a single compute zone that you specify at the time of creation. You can improve your clusters' availability and resilience by creating regional clusters.
A regional cluster provides a single static endpoint for the entire cluster, providing you with the ability to access the cluster's control plane regardless of any outage or downtime in an individual zone.
Regional clusters distribute Kubernetes resources across multiple zones within a region. Regional cluster create three cluster masters across three zones and, by default, create nodes in three zones, or in as many zones as desired.
You can choose to create a regional cluster at the time of cluster creation. However, you cannot migrate or downgrade a regional cluster to a zonal cluster, and vice versa.
- For regional clusters that run
GPUs, there is currently
no region which has any GPU type in three zones. If you want to run GPUs in
a regional cluster, you need to specify zones using the
Attempting to create a GPU cluster that spans three zones returns an error like the following:
ERROR: (gcloud.container.clusters.create) ResponseError: code=400, message= (1) accelerator type "nvidia-tesla-k80" does not exist in zone us-west1-c. (2) accelerator type "nvidia-tesla-k80" does not exist in zone us-west1-a.
- You can't create node pools in zones outside of the cluster's zones. However, you can change a cluster's zones, which causes all new and existing nodes to span those zones.
Regional clusters are offered at no additional charge.
Using regional clusters requires more of your project's regional quotas.
Ensure that you understand your quotas and the
Kubernetes Engine pricing before using regional clusters. If you encounter an
Insufficient regional quota to satisfy request for resource error, your
request exceeds your available quota in the current region.
Additionally, you are charged for node-to-node traffic across zones. For example, if you had a service in one zone that needed to talk to a service in another zone, you would be charged for the cross-zone network traffic. For more information, refer to the "Egress between zones in the same region (per GB)" pricing on the Compute Engine pricing page.
How regional clusters work
Regional clusters replicate cluster masters and nodes across multiple zones
within in a single region. For example, a regional cluster in the
us-east1 region creates masters and nodes in all three
us-east1-d. This ensures higher availability
of resources and protects clusters from zonal downtime, as regional clusters and
their resources do not fail if a single zone fails. In the event of an
infrastructure outage, the regional control plane remains available and nodes
can be rebalanced manually or using the cluster autoscaler.
There are several benefits to using regional clusters, including:
- Resilience from single zone failure. Because regional clusters are available across a region rather than a single zone within a region, if a single zone goes down, your Kubernetes control plane and your resources are not impacted.
- Zero downtime master upgrades and reduced downtime from master failures. By using a high availability control plane, your control plane’s availability is unaffected, even during upgrades.
Persistent storage in regional clusters
Persistent storage disks are zonal resources. When you add persistent storage to your cluster, unless a zone is specified, Kubernetes Engine assigns the disk to a single zone. Kubernetes Engine chooses the zone at random. When using a StatefulSet, the provisioned persistent disks for each replica are spread across zones.
Once a persistent disk is provisioned, any Pods referencing the disk are scheduled to the same zone as the disk.
A read-write persistent disk cannot be attached to multiple nodes.
Autoscaling regional clusters
You can use the cluster autoscaler to automatically scale regional clusters. The following sections offer some considerations for using the cluster autoscaler with regional clusters.
Overprovisioning scaling limits
To maintain capacity in the unlikely event of zonal failure, you can overprovision your scaling limits.
For example, if you overprovision a three-zone cluster by 150%, you can ensure that 100% of traffic is routed to available zones if one-third of the cluster's capacity is lost. In the above example, you would accomplish this by specifying a maximum of six nodes per zone rather than four. If one zone fails, the cluster scales to twelve nodes in the remaining zones.
Similarly, if you overprovision a two-zone cluster by 200%, you can ensure that 100% of traffic is rerouted if half of the cluster's capacity is lost.
To learn about autoscaling limits for regional clusters, refer to Autoscaling limits.
Balancing across zones
To learn how the cluster autoscaler balances the size of your cluster across multiple zones, refer to Balancing across zones.