kubespray/README.md

5.7 KiB

kubernetes-ansible

Install and configure a kubernetes cluster including network overlay and optionnal addons. Based on CiscoCloud work.

Requirements

Tested on Debian Jessie and Ubuntu (14.10, 15.04, 15.10). The target servers must have access to the Internet in order to pull docker imaqes. The firewalls are not managed, you'll need to implement your own rules the way you used to.

Ansible v1.9.x

Components

Ansible

Download binaries

A role allows to download required binaries which will be stored in a directory defined by the variable 'local_release_dir' (by default /tmp). Please ensure that you have enough disk space there (about 1G).

Note: Whenever you'll need to change the version of a software, you'll have to erase the content of this directory.

Variables

The main variables to change are located in the directory environments/[env_name]/group_vars/k8s-cluster.yml.

Playbook

---
- hosts: downloader
  sudo: no
  roles:
    - { role: download, tags: download }

- hosts: k8s-cluster
  roles:
    - { role: etcd, tags: etcd }
    - { role: docker, tags: docker }
    - { role: overlay_network, tags: ['calico', 'flannel', 'network'] }
    - { role: dnsmasq, tags: dnsmasq }

- hosts: kube-master
  roles:
    - { role: kubernetes/master, tags: master }
    - { role: addons, tags: addons }

- hosts: kube-node
  roles:
    - { role: kubernetes/node, tags: node }

Run

It is possible to define variables for different environments. For instance, in order to deploy the cluster on 'dev' environment run the following command.

ansible-playbook -i environments/dev/inventory cluster.yml -u root

Kubernetes

Network Overlay

You can choose between 2 network overlays. Only one must be chosen.

  • flannel: gre/vxlan (layer 2) networking. (official docs)

  • calico: bgp (layer 3) networking. (official docs)

The choice is defined with the variable 'overlay_network_plugin'

Expose a service

There are several loadbalancing solutions. The ones i found suitable for kubernetes are Vulcand and Haproxy

My cluster is working with haproxy and kubernetes services are configured with the loadbalancing type 'nodePort'. eg: each node opens the same tcp port and forwards the traffic to the target pod wherever it is located.

Then Haproxy can be configured to request kubernetes's api in order to loadbalance on the proper tcp port on the nodes.

Please refer to the proper kubernetes documentation on Services

Check cluster status

Kubernetes components

Master processes : kube-apiserver, kube-scheduler, kube-controller, kube-proxy Nodes processes : kubelet, kube-proxy, [calico-node|flanneld]

  • Check the status of the processes
systemctl status [process_name]
  • Check the logs
journalctl -ae -u [process_name]
  • Check the NAT rules
iptables -nLv -t nat

Available addons

By default 3 addons are enabled

  • A dns server in order to resolve kubernetes services names
  • Kube-ui which is a simple dashboard which shows kubernete's components, url : http://[master_ip]:8080/ui
  • Fabric8, console management for kubernetes : http://[master_ip]:8080/api/v1/proxy/namespaces/default/services/fabric8

Other addons : logging, monitoring

Calico networking

Check if the calico-node container is running

docker ps | grep calico

The calicoctl command allows to check the status of the network workloads.

  • Check the status of Calico nodes
calicoctl status
  • Show the configured network subnet for containers
calicoctl pool show
  • Show the workloads (ip addresses of containers and their located)
calicoctl endpoint show --detail

Flannel networking

Test the dns server

  • Create a file 'busybox.yaml' with the following content
apiVersion: v1
kind: Pod
metadata:
  name: busybox
  namespace: default
spec:
  containers:
  - image: busybox
    command:
      - sleep
      - "3600"
    imagePullPolicy: IfNotPresent
    name: busybox
  restartPolicy: Always
  • Create the pod
kubectl create -f busybox.yaml
  • When the pod is ready, execute the following command
kubectl exec busybox -- nslookup kubernetes.default

You should get an answer from the configured dns server

Congrats ! now you can walk through kubernetes basics

Known issues

Node reboot and Calico

There is a major issue with calico-kubernetes version 0.5.1 and kubernetes prior to 1.1 : After host reboot, the pods networking are not configured again, they are started without any network configuration. This issue will be fixed when kubernetes 1.1 will be released as described in this issue

Monitoring addon

Until now i didn't managed to get the monitoring addon working.

Listen on secure port only

Currently the api-server listens on both secure and insecure ports. The insecure port is mainly used for calico. Will be fixed soon.

Author Information

Smana - Smaine Kahlouch (smainklh@gmail.com)