Skip to content

Latest commit

 

History

History
132 lines (94 loc) · 4.69 KB

AWS.md

File metadata and controls

132 lines (94 loc) · 4.69 KB
Raw

#Kubernetes on Amazon EC2 cloud

The difference between installation using README.md and this guide is that with this guide the installed Kubernetes knows it's running under AWS and e.g. downloads list of nodes from it instead of using static configuration file. Using README.md is perfectly fine if you don't need such features.

Unlike official cluster/kube-up.sh script, which creates AWS networks, subnets, routing tables, security groups, keys and whatnot from scratch (and messes up your AWS configuration pretty much), this playbook installs and configures Kubernetes on already running instances. You should set up your account accordingly before. This playbook does not take any assumption about your AWS networking and security, use whatever you find appropriate. Even the defaults should be fine.

Prerequisites

  1. Create your AWS instances as usual, using any networking and settings. Make sure these conditions are met:

    • The instances can 'see' each other and there is no firewall between them. Of course, you should set up firewall between the instances and the Internet.

    • The instances are based on a Fedora image (Fedora 21 was tested).

    • kubernetes-0.12 or newer is available through yum or dnf. At the time of writing this document, it was necessary to enable updates-testing repository:

      $ yum install yum-utils
$ yum-config-manager --enable updates-testing

    • Choose an unique name of your Kubernetes cluster and add following tags to your AWS instances:

      Master:

      Role=<clustername>-master
Name=<clustername>-master

      Minions, where N is a sequence number:

      Role=<clustername>-minion
Name=<clustername>-minion-<N>

  2. Edit group_vars/all.yml and configure your remote ssh account name. It's usually 'fedora':

     ansible_ssh_user: fedora

  3. Edit group_vars/all.yml and configure your cluster name. The name must be the same as the <clustername> used for tags above!

     kube_aws_cluster_name: <clustername>

  4. Depending on your network setup inside AWS, you may need to set up proper /etc/hosts. Make sure your instances can resolve their internal hostnames (such as ip-10-3-88-220.ec2.internal) and ping each other using them. If they cannot, this playbook can create appropriate /etc/hosts for you. Just enable hosts setup in group_vars/all.yml:

     kube_create_etc_hosts: true

  5. Configure your AWS CLI according to http://docs.aws.amazon.com/cli/latest/userguide/cli-chap-getting-started.html. What we really need are ~/.aws/config and ~/.aws/credentials files. The playbook will copy these files to /etc/aws.conf and /etc/aws/credentials on your Kubernetes master in order to download list of minios from AWS instead of a static configuration file.

Network-specific configuration

This playbook can run either inside AWS (i.e. on an AWS instance with installed ansible) or outside it (i.e. on a machine in your home / company network).

Outside AWS

Follow these steps if you want to run this playbook on a machine outside AWS network, for example on your home or company computer. As AWS assigns public IP addresses of your instances dynamically, you have to edit inventory file every time you (re)start your AWS instances.

  1. Set up your SSH so /bin/ssh does not need any special arguments to access your instances. For example, my ~/.ssh/config looks like:

     Host *.amazonaws.com
 IdentityFile ~/.ssh/my-aws-key.pem

  2. Add external hostnames of your instances into inventory file. These hostnames are not stable and change with each reboot!

     [masters]
 ec2-52-1-61-250.compute-1.amazonaws.com

 [etcd]
 ec2-52-1-61-250.compute-1.amazonaws.com

 [minions]
 ec2-52-1-61-251.compute-1.amazonaws.com
 ec2-52-1-61-252.compute-1.amazonaws.com

Inside AWS

Follow these steps if you have an AWS instance with Ansible and you want to run this playbook there. This way, your don't need to edit your inventory file all the time.

  1. Set up you ~/.ssh/config in similar way:

    Host *
IdentityFile ~/.ssh/my-aws-key.pem

  2. Add internal IP addresses of your instances into the inventory file. These IP addresses are stable.

     [masters]
 10.3.88.220

 [etcd]
 10.3.88.220

 [minions]
 10.3.88.221
 10.3.88.222

Deploy

When everything is ready, you can finally run the playbook:

$ ansible-playbook -i inventory setup.yml.

Sometimes, reboot helps to start everything in the right orded. To test your setup, check that kubernetes knows all mininons:

$ kubectl get nodes