undercloud-live

Tools and scripts to build an undercloud Live CD and configure an already running Fedora 19 x86_64 system into an undercloud. The script is meant to be run on physical hardware. However, it can also be used on a vm. When using a vm you need to make sure that the vm you intend to configure as a undercloud has been configured to use nested kvm (see References).

To get started, clone this repo to your home directory:

$ cd
$ git clone https://github.com/slagle/undercloud-live.git

bin/undercloud.sh

This script is run as the current user to configure the current system into an undercloud.

The undercloud makes use of the default libvirtd network of 192.168.122.0/24. If you want to change the network (e.g., you're running the script on a vm whose host is already using 192.168.122.0/24), edit undercloud-live/bin/custom.sh, and then source that file:

# Edit undercloud-live/bin/custom-network.sh, and set the environment
# variables in the file to your desired settings.
$ vi undercloud-live/bin/custom-network.sh
$ source undercloud-live/bin/custom-network.sh

Run the undercloud script itself:

$ undercloud-live/bin/undercloud.sh

The script logs to ~/.undercloud-live/undercloud.log. If there is an error applying one of the diskimage-builder elements, you will see a prompt to continue or not. This is for debugging purposes.

Once the script has completed, you should have a functioning undercloud. At this point, you would move onto the next steps of building images and deploying an overcloud. These steps are also scripted in the images.sh and deploy-overcloud.sh scripts. You can just run these scripts if you prefer to do that instead:

$ undercloud-live/bin/images.sh
$ undercloud-live/bin/deploy-overcloud.sh

NOTE: undercloud-images.sh will not build images if the files already exist under /opt/stack/images. If you already have image files you want to use on the undercloud, just copy them into /opt/stack/images.

Prerequisites

• Only works on Fedora 19 x86_64
• sudo as root ability

Caveats

• undercloud.sh deploys software from git repositories and directly from PyPi. This will be updated to use rpm's at a later date.
• The git repositories that are checked out under /opt/stack are set to checkout specific hashes. Some of these hashes are specified in bin/install.sh. Others are specified in an undercloud-live branch of a fork of tripleo-image-elements at https://github.com/slagle/tripleo-image-elements.git. The undercloud-live branch there sets specific hashes to use via the source-repository interface.
• If you reboot the undercloud system, you will need to rerun bin/network.sh
• The system is configured to use the iptables service instead of the firewalld service.
• SELinux is set to Permissive mode. Otherwise, rabbitmq-server will not start.
  See: https://bugzilla.redhat.com/show_bug.cgi?id=998682
  Note: we will be switching to use qpid soon

kickstart/fedora-undercloud-livecd.ks

kickstart file that can be used to build an undercloud Live CD.

1. install fedora-kickstarts and livecd-tools if needed
2. livecd-creator --debug --verbose --fslabel=Fedora-Undercloud-LiveCD --cache=/var/cache/yum/x86_64/19 --releasever=19 --config=/path/to/undercloud-live/kickstart/fedora-undercloud-livecd.ks

This will produce a Fedora-Undercloud-LiveCD.iso file in the current directory. To test it simply run:

qemu-kvm -m 2048 Fedora-Undercloud-LiveCD.iso

(you can run it with less ram, but it will be quite a bit slower)

Live CD

The Live CD provides a full working undercloud environment. Note that the Caveats from the section above apply to the Live CD as well as it is built using the same installation scripts.

Also keep in mind that any changes made to the filesystem while running the Live CD are lost after you reboot or use the install to disk method. Therefore, if you plan to install, I recommend doing this first before building images on your undercloud, etc.

Currently the Live CD uses an XFCE based desktop for no particular reason other than that it builds faster, uses less disk space, and has a much more responsive desktop when testing from within a vm. It can be switched to Gnome later.

Requirements

1. RAM

• = 8GB if you plan to immediately install the live cd to disk

• = 16GB if you plan to just run the live cd and immediately start building images, etc.

1. Disk (if you plan to install)

• = 25GB (the bigger the better obviously if you plan to upload/build many images)

1. Nested KVM (if you plan to use a vm for the Live CD itself)

• setup up Nested KVM (see References below)
• if you don't want to use Nested KVM, make sure you switch all your vm's to use just qemu virtualization in their libvirt xml.

Running/Installing

To use the live cd, follow the steps below.

1. Boot the live cd.

• The default account is liveuser. However, you can use stack/stack for ssh access, etc.
• If you plan to install to disk, do so after the boot is finished. Use the icon on the desktop, or ssh in with X forwarding and run /sbin/liveinst.
• Once the install has finished, reboot and continue on with the next step. After rebooting, you will need to use stack/stack to login as liveuser no longer exists.

1. Open a terminal and switch to the stack user (if you aren't already):

    su -
    su - stack
   

2. The libvirtd service on the undercloud uses the default network of 192.168.122.0/24. If this is already in use in your enviornment, you need to change it. Here's an example of doing that if you wanted to switch the subnet to 123 instead of 122:

    # run these commands as root
    sed -i "s/122/123/g" /etc/libvirt/qemu/networks/default.xml
    systemctl restart libvirtd
    virsh net-destroy default
    virsh net-start default
    # Update nova compute configuration
    sed -i "s/192.168.122.1/192.168.123.1/g" /etc/nova/nova.conf
    systemctl nova-compute restart
    # Update embedded heat meatdata (it gets applied on every boot)
    sed -i "s/192.168.122.1/192.168.123.1/g" /var/lib/heat-cfntools/cfn-init-data
   

3. Source the undercloud configuration

    source undercloudrc
   

From here, you can use all the normal openstack clients to interact with the running undercloud services.

To get going on deploying an overcloud, you will want to build images and start the overcloud. There are scripts to do these pieces as well, but we may change that into just documentation instructions so that users get the full experience of setting up an overcloud themselves. The scripts are here:

/opt/stack/undercloud-live/bin/images.sh
/opt/stack/undercloud-live/bin/deploy-overcloud.sh

After the overcloud is deployed, you can do the following to interact with it's services:

export OVERCLOUD_IP=$(nova list | grep notcompute.*ctlplane | sed  -e "s/.*=\\([0-9.]*\\).*/\1/")
source /opt/stack/tripleo-incubator/overcloudrc

A couple of points to remember as the Live CD is used:

1. You can use the Install to Hard Drive shortcut on the desktop to install the live cd to disk. When you do this, any changes that you had made, will be lost and you'll need to start over with image building, etc when you boot the now installed undercloud.
2. When running the live cd there is only 512mb worth of changes that can be applied to the root filesystem and this fills up rather quickly with just logs, etc. As such, the following directories are all tmpfs mounted:
3. /home/stack/.cache/image-create/ccache
4. /home/stack/.cache/image-create/yum
5. /opt/stack/images
6. /var/lib/glance/images
7. /var/lib/libvirt/images
8. /var/lib/nova/instances

References

Nested KVM Setup Help

1. http://www.server-world.info/en/note?os=Fedora_19&p=kvm&f=8
2. https://fedoraproject.org/wiki/QA:Testcase_KVM_nested_virt