Open kevinayres opened 11 months ago
This works for creating the instance, EIP, association:
provider "aws" { profile = "default" }
variable "ami-sles15sp5payg" { type = string default = "ami-0aa3dc9f3f70b91a7" #AMI for us-west-1 }
variable "key-name" { type = string default = "access" }
resource "aws_instance" "demo-instance" { ami = var.ami-sles15sp5payg instance_type = "t2.micro" key_name = var.key-name }
resource "aws_eip" "demo-1" { vpc = true }
resource "aws_eip" "demo-2" { instance = aws_instance.demo-instance.id vpc = true } `
I ran into this problem as well. I shut down the two instances to save money over the weekend, and when they powered up again, they received different IP addresses. terraform destroy
now fails with:
╷
│ Error: Kubernetes cluster unreachable: Get "https://13.38.67.225:6443/version": dial tcp 13.38.67.225:6443: i/o timeout
│
│ with module.rancher_common.helm_release.cert_manager,
│ on ../rancher-common/helm.tf line 4, in resource "helm_release" "cert_manager":
│ 4: resource "helm_release" "cert_manager" {
│
╵
Now I have to manually dig around and destroy the infrastructure I was using to evaluate Rancher. It would be great if the quickstart could be refactored to get the current IP before attempting helm operations in Terraform, or just use an EIP as suggested above.
Issue: By using a public IP (not EIP), things like Nginx are bound to that temporary IP. If you shutdown/start the instance, you get a different IP resulting in 'bad gateway' when accessing the API.
One way this project could be even more helpful would be to support EIP on deployment. Example: