Skip to content

Latest commit

 

History

History
163 lines (116 loc) · 5.68 KB

HOWTO.md

File metadata and controls

163 lines (116 loc) · 5.68 KB
Raw

How to automatically deploy SUSE OpenStack Cloud via Vagrant

Prerequisites

Please see the prerequisites page for information on hardware requirements and how to set up Vagrant to work with your hypervisor.

SUSE OpenStack Cloud installation

N.B. The following steps describe semi-automatic booting of the cloud infrastructure via Vagrant. Another more fully automated option is to use one of the pre-canned demos.

  • Tell Vagrant which provider you want to use, e.g.

    export VAGRANT_DEFAULT_PROVIDER=libvirt

    With Vagrant 1.6.x, VirtualBox is the default; although future versions (i.e. 1.7.x) will instead automatically detect the best default provider.

  • Depending on what cloud configuration you desire, either use Vagrant to sequentially provision all four VMs from the default configuration (1 admin + 2 controllers + 1 compute node) in one go:

    cd vagrant
vagrant up

    or keep reading to find out how to choose which VMs to provision.

There is always exactly one admin server node. The quantity, shape, and size of all nodes are determined by a YAML config file. The default is 2-controllers-1-compute.yaml but there are other examples in the same directory.

You can change the number of controller nodes and compute notes from the defaults of 2 and 1 respectively by editing this file or by pointing the Vagrantfile at an alternative config file:

export VAGRANT_CONFIG_FILE=/path/to/other/vagrant.yaml

vagrant up --no-parallel will cause all the VMs to be provisioned in the order listed in the YAML config file. Typically this is:

  1. admin - the Crowbar admin server node. After boot-up, install-suse-cloud will automatically run. This takes quite a few minutes to complete, since it has to start several services. Once you see the next VM start to boot, you know it has completed installation, at which point you can visit the Crowbar web UI (username and password are both crowbar by default) and watch the other nodes come online one by one.
  2. The controller node(s) in numerical order: controller1, then controller2 etc. These will run the OpenStack infrastructure services, typically within a Pacemaker cluster.
  3. The compute nodes in numerical order: compute1, then compute2 etc.

It will take a few minutes to provision each VM, since not only does Vagrant need to copy a fresh virtual disk from the box for each VM, but also on first boot the VMs will register against Crowbar and then perform some orchestrated setup via Chef.

Alternatively, you can provision each VM individually, e.g.

vagrant up admin
vagrant up controller1

and so on. Similarly, you can rebuild an individual VM from scratch in the normal Vagrant way, e.g.

vagrant destroy compute1
vagrant up compute1

or the entire cloud in one go:

vagrant destroy
vagrant up --no-parallel

CAUTION! If you are using libvirt, please see this page about usage vagrant-libvirt for known pitfalls.

Connecting to the VMs

Of course you can log in on the VM consoles, via the hypervisor's GUI.

You can also ssh via vagrant, e.g.:

vagrant ssh admin
vagrant ssh controller1 -- -l root

or directly to the admin node:

ssh [email protected]

The root password is vagrant, as per convention.

Setting up node aliases

By default, Crowbar and Chef name nodes according to the MAC address of their primary interface. This is not very human-friendly, so Crowbar offers the option of assigning aliases (e.g. controller1, compute1 etc.) to nodes. This Vagrant environment provides a simple script to automate that: once you have booted your controller and compute nodes, simply ssh to the admin server as per above, and run:

setup-node-aliases.sh

After you have done this, the admin server's DNS tables will update, and you will be able to ssh conveniently from the admin node to other nodes, e.g.

ssh controller1

etc.

N.B. The ./build.sh demo scripts take care of this automatically.

Setting up node/role shell variables (OPTIONAL)

If you want to poke around behind the scenes and see how Crowbar uses Chef, you can use the knife command on the admin server. However it can quickly get tedious figuring out the correct node or role name corresponding to say, controller2. Therefore the Vagrantfile automatically installs a handy node-sh-vars script which can set convenient shell variables for the node and role names. Once you have booted your controller and compute nodes, simply ssh to the admin server as per above, and run:

node-sh-vars > /tmp/.crowbar-nodes-roles.cache
source /tmp/.crowbar-nodes-roles.cache

Now you can easily look at Chef nodes via knife, e.g.

knife node show $controller1
knife node list

and similarly for roles:

knife role show $controller1r
knife role list

The Vagrantfile also sets up .bash_profile so that /tmp/.crowbar-nodes-roles.cache is automatically sourced on login.

N.B. The ./build.sh demo scripts take care of this automatically.

Trying out SUSE OpenStack Cloud