StarlingX Installer

StarlingX Installer is a collection of tools designed for automating installation of StarlingX images on VirtualBox or LibVirt instances according to official instructions that can be found here (for an All-in-One system):

https://wiki.openstack.org/w/index.php?title=StarlingX/Containers/Installation&oldid=170746

Description

This installer is actually a factory of installers.

Given the following inputs:

• the type of setup to be installed:
  • aiosx: all-in-one with one server (all services running on the same machine)
  • aiodx: all-in-one with two servers (every controller is also a compute)
  • standard: two controller nodes and two or more computes
  • storage: two controller nodes, two or more storage nodes, two or more computes
• configuration specific to the selected lab:
  • OAM network (floating IP, controller IP addresses, etc.)
  • VirtualBox NAT network
  • CPUs, memory and disk sizes for each node type
  • number of compute and storage nodes, etc.

The factory will output an archive containing shell scripts to:

• setup VirtualBox or LibVirt VMs and networking
• patch ISO with configuration needed by the automated installer
• automatically install a StarlingX release

This approach makes the install process more transparent because generated shell scripts are readable and easy to update in case that's needed.

Usage

To start the factory web server:

stxlab

Will start an HTTP server on port 3000. The port number can be changed using command line option -port. Main page contains 4 tabs, one for each setup type: aiosx, aiodx, standard and storage. After updating configuration options (already populated with defaults) user can click Generate ... Installer button to download an archive containing installer scripts.

The web server is just a simple GUI front-end to access functionality provided by installer-generator methods. The same functionality can be accessed via CLI using binaries provided for each setup type: aiosx, aiodx, standard, storage.

Each binary has a large set of command line options to fine tune VirtualBox or Libvirt VMs and the installed StarlingX. All of them have default values so user can run the commands without providing any option.

As with the web server, the result is an archive containing installer scripts. This enables scenarios like:

• local deployment of install scripts from local factory:

  ./aiosx | tar -C /path/to/lab/workspace -x
  

• remote deployment of install scripts from local factory:

  ./aiosx | ssh user@lab-machine tar -C /path/to/lab/workspace -x

• local deployment of install scripts from remote factory:

  ssh user@installer-factory /path/to/aiosx | tar -C /path/to/lab/workspace -x
  

The generated archive contains:

• setup.sh: create VirtualBox or LibVirt VMs
• prepare-bootimage.sh: prepare StarlingX ISO for automatic install
• install.sh: automatic installer
• vmctl.sh: snapshot/restore the entire lab

In order to patch the StarlingX ISO the following utilities need to be available before running the VM setup script:

fuseiso rsync fusermount mkisofs

On Ubuntu they can be installed by running:

sudo apt install fuseiso rsync fuse genisoimage

Install steps are the following:

1. unpack installer archive. Assuming an all-in-one single node setup the instructions are:

   aiosx | tar -x
   cd aiosx
   

   (to install using LibVirt run aiosx --hypervisor libvirt)

2. create VMs:

   ./setup.sh /path/to/starlingx/iso
   

   This will remove already existing VMs named <setup-type>-controller-0, etc. and create a new set of VMs according to specification then group them under <setup-type> (where setup-type can be aiosx, aiodx, standard, storage).

   Then setup.sh calls prepare-bootimage.sh to patch the provided ISO image with the lab configuration and to make the automated install process independent of console access via serial port. The patched ISO goes through CentOS Anaconda steps then reboots to a state where it is accessible via SSH.

3. (optional) provide OpenStack charts:

   cp /path/to/stx-openstack/chart.tgz ./stx-openstack.tgz
   

   Note that the name of the file must match exactly stx-openstack.tgz. Installer applies platform-integ-apps then exits if it can't find this file.

4. run automated installer:

   ./install.sh
   

   Which will go through the steps of running ansible, adding, configuring and unlocking additional nodes, waiting for platform-integ-apps then applying stx-openstack (when available).

   If the install procedure crashes at some point and you need to resume it from a specific step there's a shell here-doc marker SKIP_SECTION that can be moved to include the section of the file you need to skip (meaning all steps that were already executed).

vmctl.sh is provided in case you need to snapshot or restore the entire setup. It pauses all the VMs before taking a snapshot and restores them to a paused state then quickly resumes all of them to minimize clock skew between VMs.

setup.sh creates an env.sh that can be sourced to enable shell aliases for serial console access (they map to socat commands). For example:

 source ./env.sh
 aiosx-controller-0-serial

install.sh creates a local ssh.config file that can be used to access installed nodes, for example ssh -F ssh.config controller-0

Build

Clone this repository and run make to build it:

make

Will create in ./bin binaries that are self contained (include also template files, thanks to packr) and can be moved/deployed somewhere else.