Closed liangruibupt closed 5 years ago
! update ! The root cause of 'dial tcp 172.16.101.115:443: i/o timeout' is my bastion server (run make validate-cluster) is in different VPC of k8s VPC. I need setup VPC peering or create new bastion server on k8s VPC.
BTW, some times, some ELB instances are not InService more than 15 mins, you can make delete-cluster and re-create cluster again.
Makefile 可以参考issue 里面的样例,其就是就是添加3行 --utility-subnets=$(SUBNET_IDS) \ --api-loadbalancer-type=internal \ --topology=private
@liangruibupt 我看到代码中有SUBNET_IDS ?=subnet-01f7afb87dd26bc9c,subnet-0be82b17cbe107b98
,所以这样的话是需要先自己创建subnet,然后指定给kops用吗,而不是跟之前一样让kops帮我们创建subnet?
Kops應該不會主動建private subnets以及NAT,因此 @liangruibupt 提供的範例應該是自己建立的private subnets然後當作arguments帶給Kops使用。
Put the master and nodes to private subnet with NAT gateway.
Makefile snippet: using existed VPC and subnet, AMI is AmazonLinux2, follow up the https://github.com/kubernetes/kops/issues/4548 to add --utility-subnets=$(SUBNET_IDS) --api-loadbalancer-type=internal --topology=private to create_cluster
keep other parts of Makefile no change
validate-cluster failed with below error:
Already checked the issue, https://github.com/nwcdlabs/kops-cn/issues/5