The purpose of the simple
kubernetes_cluster_vmware repo is to provide simple scripts for building a CoreOS cluster using the methodology that Kelsey Hightower gave me insight into using the official VMware CoreOS image. I modified this process along with the information in the Kubernetes project for provisioning a Kubernetes with a CoreOS Multi-node cluster.
This project builds on the coreos_cluster_vmware project, which is a good starting point to get a basic understanding for the process of putting together this Kubernetes cluster using the same essential process.
What does this cluster consist of?
Four machines total:
- 1 master (with gui)
- 3 minions (headless)
What is in this repo?
In the base directory, there are "templates" for:
vm_tmpl.vmx- The VMX file used by all VMs
master.yaml- A cloud init file for the master VM
node.yaml- A cloud init file for the minion VM nodes
get_image.sh- Simple script to fetch the official VMware CoreOS image and un-compress it (bzip)
build_master.sh- The script used to build configuration files for and to launch the master VM
build_nodes.sh- The script that is used to build configuration files for and to launch all three of the minion VMs
teardown_nodes.sh- Stops and deletes the files for minion VMs
teardown_master.sh- Stops and deletes the files for the master VM
start_master.sh- Starts the master after it has been stopped
start_nodes.sh- Starts the nodes after they has been stopped
The basic idea is this - - Obtain the official VMware CoreOS image - Produce both a cloud init file and VMX file for a given VM - Create a config drive, using the cloud init file (.iso) - Make a copy of the official VMware CoreOS image for that machine named accordingly - Boot the VM using the generated files
For the master VM, this only happens once and requires no cognizance of any other machines. For each minion VM, they boot the same way but also need to know the IP address of the master VM.
Upon launching all minion VMs, everything should be up and running!
There is a directory in the repo with nothing in it
work_dir. Enter that directory. This is where all the generated files and VMware images will exist.
Determine location of VMware CLI
You will need to find the utility
vmlist. On OSX Yosemite, this location should be
/Applications/VMware Fusion.app/Contents/Library. Set up the $PATH environment variable to have this in your path:
export PATH=$PATH:/Applications/VMware\ Fusion.app/Contents/Library
Get the official VMware CoreOS image
After this script is completed, there should be the image in the expected location
reason:work_dir patg$ ls -l ../coreos_image/*.vmdk total 1036072 -rw-r--r-- 1 patg staff 396820480 Mar 12 12:59 coreos_production_vmware_image.vmdk
Lanch the master VM
reason:work_dir patg$ ../bin/build_master.sh Creating hybrid image... ....
This will lanch the master VM. You can then find out what the IP address of the Virtual Machine by running the following command:
reason:work_dir patg$ vmrun getGuestIPAddress master.vmx 172.16.230.132
Log into the instance. The password that was set from the cloud init data file
master_cloud_init.yaml results in the VM having a password for both the core and root user of "vmware" (NOTE: this is not for production, obviously, so you will want to change it prior!)
reason:work_dir patg$ ssh email@example.com Warning: Permanently added '172.16.230.132' (RSA) to the list of known hosts. firstname.lastname@example.org's password: CoreOS alpha (618.0.0)
Now, verify that etcd is running:
core@master ~ $ etcdctl ls /coreos.com /registry
Launch the minions
Now the minions can be launched. As the above example shows, the IP address for master (etcd) is 172.16.230.132. This will be the single argument to the next script:
reason:work_dir patg$ ../bin/build_nodes.sh 172.16.230.132
This will result in the same sequence of steps as the master, but 3 times. Once all VMs are launched, you can verify that they are up:
reason:work_dir patg$ sudo vmrun list Total running VMs: 4 /Users/patg/code/kubernetes_cluster_vmware/work_dir/node_01.vmx /Users/patg/code/kubernetes_cluster_vmware/work_dir/node_02.vmx /Users/patg/code/kubernetes_cluster_vmware/work_dir/master.vmx /Users/patg/code/kubernetes_cluster_vmware/work_dir/node_03.vmx
Now that everything is running, copy
kubectl to the master VM:
reason:work_dir patg$ scp ../kubectl email@example.com:~/ firstname.lastname@example.org's password: kubectl 100% 13MB 13.2MB/s 00:00
Log into the master and copy
reason:work_dir patg$ ssh email@example.com firstname.lastname@example.org's password: CoreOS alpha (618.0.0) core@master ~ $ sudo mv kubectl /opt/bin
Test that everything is working with the base coreos setup (using
core@node_03 ~ $ fleetctl --endpoint=http://172.16.230.132:4001 list-machines MACHINE IP METADATA 0bd560c9... 172.16.230.132 role=master 10a1d3c3... 172.16.230.133 role=node 50a0dd1c... 172.16.230.134 role=node 9bd07cf8... 172.16.230.135 role=node
Excellent! A working coreos cluster! This also will give you the IP addresses of the minions - make a note.
Now, verify that Kubernetes is working, and list the minions:
core@master ~ $ kubectl get minions NAME LABELS STATUS 172.16.230.134 <none> Ready 172.16.230.133 <none> Ready 172.16.230.135 <none> Ready
The Kubernetes cluster is now open for business!
As good start is the excellent example for redis that comes with the Kubernetes source.
Clone kubernetes on the master and run the examples (read the documents for more information):
core@master ~ $ git clone https://github.com/GoogleCloudPlatform/kubernetes.git Cloning into 'kubernetes'... remote: Counting objects: 59225, done. remote: Total 59225 (delta 0), reused 0 (delta 0), pack-reused 59225 Receiving objects: 100% (59225/59225), 36.64 MiB | 747.00 KiB/s, done. Resolving deltas: 100% (38651/38651), done. Checking connectivity... done. core@master ~/kubernetes $ cd kubernetes/examples/redis/ core@master ~/kubernetes/examples/redis $ kubectl create -f redis-master.yaml redis-master core@master ~/kubernetes/examples/redis $ kubectl create -f redis-sentinel-service.yaml redis-sentinel core@master ~/kubernetes/examples/redis $ kubectl create -f redis-controller.yaml redis core@master ~/kubernetes/examples/redis $ kubectl create -f redis-sentinel-controller.yaml redis-sentinel core@master ~/kubernetes/examples/redis $ kubectl get services NAME LABELS SELECTOR IP PORT kubernetes component=apiserver,provider=kubernetes <none> 10.100.0.2 443 kubernetes-ro component=apiserver,provider=kubernetes <none> 10.100.0.1 80 redis-sentinel name=sentinel,role=service redis-sentinel=true 10.100.9.218 26379 core@master ~/kubernetes/examples/redis $ kubectl get pods POD IP CONTAINER(S) IMAGE(S) HOST LABELS STATUS redis-master 10.244.89.2 master kubernetes/redis:v1 172.16.230.134/172.16.230.134 name=redis,redis-sentinel=true,role=master Pending sentinel kubernetes/redis:v1 core@master ~/kubernetes/examples/redis $ kubectl get rc CONTROLLER CONTAINER(S) IMAGE(S) SELECTOR REPLICAS redis redis kubernetes/redis:v1 name=redis 1 redis-sentinel sentinel kubernetes/redis:v1 redis-sentinel=true 1 core@master ~/kubernetes/examples/redis $
This document has shown how to easily set up a Kubernetes cluster, as well as how to do some useful work with the VMware command line tools. For more information, do join the
#google-containers IRC channels on Freenode, as well as the documentation in the Kubernetes project.
Lastly, many many thanks to Kelsey Hightower and others for their patience and help with setting both stock CoreoOS and Kubernetes and answering a slew of questions!