IMPORTANT! This article has been deprecated. For current information about using Cumulus Linux with OpenStack, please read our OpenStack Solutions page.
This article discusses how you can easily implement the configuration discussed in the OpenStack and Cumulus Networks Validated Design Guide in the Cumulus Workbench. This demo illustrates the deployment of OpenStack on network switches running Cumulus Linux. Plus it emphasizes the use of CLAG (Chassis Link Aggregation) in a Layer 2 environment.
This demo package uses Puppet to quickly run through almost all of the manual configuration steps listed in the design guide. It automates steps 1-9 of the design guide; you only have to start the VMs using the OpenStack Horizon Web interface.
{{table_of_contents}}
Features Demonstrated
- Cumulus Workbench
- Openstack Validated Design Guide
- CLAG (multi-Chassis Link Aggregation)
- Prescriptive Topology Manager (PTM)
- Zero touch provisioning (ZTP)
- OpenVPN on Cumulus Workbench
Source Code
Network Diagrams
The topology has 2 spine switches, 2 leaf switches with 2 server hosts:
Files and Descriptions
File | Description |
/var/www/provision-openstack.sh | ZTP automation script to install ssh authorized keys; installs and triggers puppet |
/var/www/topology.dot | Network topology file for PTM to validate against |
/etc/puppet/modules/openstack/ | Puppet configuration files for Openstack |
/etc/puppet/manifests/site.pp | Node specific manifests for puppet |
Running the Demo
Preparing the Environment
Once you have logged into the workbench and accepted the end user license agreement, you can install Puppet code.
- Install the demo package on your workbench:
root@wbench:~# apt-get update
root@wbench:~# apt-get install cldemo-wbench-openstack-2lt22s-puppet
- Set up and run OpenVPN on your workbench.
Giving the Demo
- Restart all switches to install CLAG on all switches via Puppet. From the workbench, run:
root@wbench:~# cw-swpower -a -o reset
- Use Puppet to install Ubuntu Server and OpenStack Horizon on each server host from the workbench. This step takes about 25 minutes:
root@wbench:~# cw-pxehelper -s server1 -o ubuntuserver-trusty-puppet -n
root@wbench:~# cw-pxehelper -s server2 -o ubuntuserver-trusty-puppet -nThis step installs and configures all of the OpenStack services.
- Follow step 10 in the design guide to connect to the OpenStack Horizon Web UI and start the virtual machines. The OpenStack Horizon Web interface is at http://192.168.0.201/horizon. You connect to it via OpenVPN, which you set up earlier. The username is
admin
and the password isadminpw
. - Login, and under the project section select instances.
- Click the Launch Instance button.
- Name your instance. Under "Instance Boot Source" select "Boot from Image." For "Image Name" select Cirros. Then select "Launch."
Sometimes, the interfaces on the Ubuntu hosts do not come up properly, running 'puppet agent -- test' on the Ubuntu servers can help fix it.root@server1:~# puppet agent --test
root@server2:~# puppet agent --test
- To connect to your new instance, you can either ssh from server1 to the IP address listed (login as 'cirros' user. default password: 'cubswin:)', use 'sudo' for root), or click through the instance and use a web console.
Using SSH port forwarding
Using OpenVPN is the recommended solution, however you can also use port forwarding to access the instances. To forward, replace NNN with your workbench id (ex:406)
ssh -L 8080:192.168.0.201:80 -p 7NNN cumulus@workbench.cumulusnetworks.com
Replace all instances of http://192.168.0.201/horizon with http://localhost:8080/horizon
Repeating the Demo
To give the demo more than once, you must clear the overlay file system, clear the servers, and reboot the devices. The simplest way to do this is to just request a reprovisioning of your workbench; however, you may do this manually.
- Login to each switch and run:
cumulus@leaf1$ sudo /usr/cumulus/bin/cl-img-clear-overlay -f 1
,then reboot - Clear the server operating system:
root@wbench:/# apt-get install cldemo-wbench-osinstaller-erasembr
- Reload server1 with:
root@wbench:/# cw-pxehelper -s server1 -o erasembr -n
- Repeat the previous step for server2.
- When the switches reboot the host SSH keys will have changed, so you must remove the
known_hosts
file on the workbench VM.root@wbench:/# rm /root/.ssh/known_hosts
- You must revoke the puppet certificates on the workbench:
root@wbench:/# puppet cert clean --all
The environment is now in a state where you can repeat the demo.
Comments