The v10.0 release of kubernetes-engine is a backwards incompatible release.
- For beta modules, support for Google provider versions older than v3.29 has been removed.
- For GA modules, provider version constraints are unchanged.
The new default machine type for node pools is now e2-medium
. These VMs are more cost-effective. e2-medium machine types sustain 2 vCPUs at 50% of CPU time, yielding the same processing capacity as 1 vCPU and can temporarily burst to 2 vCPUs. Thus the sustained throughput is the same as n1-standard-1, so long as there are multiple threads running on the node -- and if you have multiple containers -- this is almost certainly the case.
However, e2-medium
nodes don't work for all use cases. In particular, if you are using any of these features you should choose another machine type:
- You're running single-threaded applications with sensitive performance requirements.
- You use GPUs or local SSDs or sole tenancy, all of which are unsupported by the new default machine type e2-medium.
If you want to keep the old node type (n1-standard-2
), you should specify it explicitly:
module "gke" {
source = "terraform-google-modules/kubernetes-engine/google"
- version = "~> 9.0"
+ version = "~> 10.0"
...
node_pools = [
{
name = "pool-01"
+ machine_type = "n1-standard-2"
},
]
}
The interface for configuring pod security policy has been simplified to a simple boolean. If you've changed the default, you'll need to update your config:
module "gke" {
source = "terraform-google-modules/kubernetes-engine/google"
- version = "~> 9.0"
+ version = "~> 10.0"
...
- pod_security_policy_config = [{
- enabled = true
- }]
+ enable_pod_security_policy = true
}