Kops-BJS

This tutorial will walk you through building a Kubernetes cluster with Kops in AWS Beijing or NinXia Region.

Name	Support Kops Version	Support Kubernetes Version	Last Update
pahud/kops-bjs	1.9.1	1.9.6	June03, 2018

Agenda

Prepare the AMI

Install Kops and Kubectl client on your laptop

Create a proxy server with gost in AWS N. Virginia Region

Create a proxy forwarder in AWS Beijing Region

Create the cluster with Kops

Prepare the AMI

Check the latest AMI ID from Kops Images document and find the AMI ID in the global regions(e.g. N. Virginia).

However, as the China Beijing region already has latest CoreOS AMI, you can just check CoreOS official EC2 AMI page and select the AMI for cn-north-1 region, make sure you select the HVM AMI type. For example, current AMI ID is ami-555a8438 (Container Linux 1745.5.0). Please note the latest AMI ID may change over time.

Region	CoreOS AMI
Beijing(cn-north-1)	ami-555a8438
NinXia(cn-northwest-1)	ami-06a0b464

Install Kops and Kubectl client on your laptop

Create a proxy server with gost and AWS Fargate

click the button to create a proxy server with gost and AWS Fargate in any of the following regions.

Region	Launch Stack in VPC	Runtime
Oregon (us-west-2)		Fargate+ECS
N. Virginia (us-east-1)		Fargate+ECS

Create a proxy forwarder in AWS Beijing or Ninxia Region

Depending on which region you woud like to provision your Kops cluster, click the button below to create an internal http_proxy forwarder for your Kops cluster. This template will create one t2.micro EC2 behind ELB in your existing VPC as the proxy forwarder.

Region	Launch Stack in VPC	Runtime
Beijing (cn-north-1)		EC2+ELB
NinXia (cn-northwest-1)		EC2+ELB

Create the cluster with Kops

update create_cluster.sh and modify the variables:

cluster_name='cluster.bjs.k8s.local'
ami='ami-xxxxxxxx'
vpcid='vpc-c1e040a5'

cluster_name : specify your cluster name. You can leave it as default. Make sure the cluster_name ends with .k8s.local so it will create gossip-based cluster without using Route53, which is not available in China Beijing region.

ami : AMI ID. See Prepare the AMI above.

vpcid: Your existing VPC ID, in which you would launch your Kubernetes cluster with Kops.

update env.config

export AWS_PROFILE='bjs'
export AWS_DEFAULT_REGION='cn-north-1'
export AWS_REGION=${AWS_DEFAULT_REGION}
export KOPS_STATE_STORE=s3://pahud-kops-state-store

AWS_PROFILE - make sure the profile name points to your AWS Beijing Region configuration. Check ~/.aws/config for details.
AWS_DEFAULT_REGION - specify cn-north-1 for Beijing Region.
KOPS_STATE_STORE - you need specify an empty S3 bucket for Kops state store, make sure you change the value and points to your S3 bucket in Beijing Region.

execute the script to create the cluster:

$ bash create_cluster.sh

edit your cluster

$ kops edit cluster cluster.bjs.k8s.local

paste the content below under the spec section for the cluster. Make sure you set correct httpProxy host.

spec:
  docker:
    logDriver: ""
    registryMirrors:
        - https://registry.docker-cn.com
  egressProxy:
    httpProxy:
      host: <host>
      port: 8888
    excludes: amazonaws.com.cn,amazonaws.cn,aliyun.cn,aliyuncs.com,registry.docker-cn.com

(you should be able to see your httpProxy host and port info in the output of the cloudformation in Beijing Region)

update the cluster with —yes

kops update cluster --name cluster.bjs.k8s.local --yes

After a few minutes(typically 8-15min), you can validate the cluster like this:

$ kops validate cluster
Using cluster from kubectl context: cluster.bjs.k8s.local

Validating cluster cluster.bjs.k8s.local

INSTANCE GROUPS
NAME			ROLE	MACHINETYPE	MIN	MAX	SUBNETS
master-cn-north-1a-1	Master	m3.medium	1	1	cn-north-1a
master-cn-north-1a-2	Master	m3.medium	1	1	cn-north-1a
master-cn-north-1b-1	Master	m3.medium	1	1	cn-north-1b
nodes			Node	m3.medium	2	2	cn-north-1a,cn-north-1b

NODE STATUS
NAME						ROLE	READY
ip-172-31-37-81.cn-north-1.compute.internal	node	True
ip-172-31-39-42.cn-north-1.compute.internal	master	True
ip-172-31-51-46.cn-north-1.compute.internal	master	True
ip-172-31-68-190.cn-north-1.compute.internal master	True
ip-172-31-68-61.cn-north-1.compute.internal	node	True

Your cluster cluster.bjs.k8s.local is ready

Or get nodes list like this

$ kubectl get nodes
NAME                                           STATUS    ROLES     AGE       VERSION
ip-172-31-37-81.cn-north-1.compute.internal    Ready     node      15m       v1.9.3
ip-172-31-39-42.cn-north-1.compute.internal    Ready     master    17m       v1.9.3
ip-172-31-51-46.cn-north-1.compute.internal    Ready     master    16m       v1.9.3
ip-172-31-68-190.cn-north-1.compute.internal   Ready     master    16m       v1.9.3
ip-172-31-68-61.cn-north-1.compute.internal    Ready     node      15m       v1.9.3

clean up

delete the cluster

$ kops delete cluster --name cluster.bjs.k8s.local --yes

And delete the two cloudformation stacks from N.Virginia and Beijing regions.

Fast Bootstrapping with local mirror

The approach provided above will not leverage any local mirror of artifacts. If you are interested to leverage local artifacts mirror including the gcr.io docker hub mirror to accelerate the boostrapping, please check this table:

Region	Guide
Beijing(cn-north-1)	fastboot guide
NinXia(cn-northwest-1)	fastboot guide

FAQ

Questions? check the FAQ list here.

WangYunzhang/kops-bjs