xp5k-openstack

Usage

From a grid5000 frontend

Initial setup

• Clone the repository :

git clone https://github.com/pmorillon/xp5k-openstack.git

Notes : Openstack delpoyment is quite long, it would be safer to continue into a screen environment.

• Setup environment :

cd xp5k-openstack
source setup_env.sh
gem install bundler
bundle install

Notes : use rehash command if you use zsh in order to update $PATH (if bundleor rakecommands are not found).

Begin an experiment

• Setup environment :

cd xp5k-openstack
source setup_env.sh

• Configure experiment. Create a file xp.conf :

site            'rennes'
walltime        '4:00:00'

public_key      "/home/pmorillo/.ssh/id_rsa.pub"
gateway         "#{ENV['USER']}@frontend.#{self[:site]}.grid5000.fr"

Notes : Ensure that you can connect to other frontends through SSH without password.

• Start experiment :

rake run

Notes : ☕️ ☕️ (estimated time: 20-30 minutes). To kill jobs before the end of the experiment, use task grid5000:clean. Before restarting a new experiment .xp_cache file should be removed.

• Connect to the controller through SSH :

rake shell host=controller

• Play with openstack

source openstack-openrc.sh
glance image-list
+--------------------------------------+----------------------+
| ID                                   | Name                 |
+--------------------------------------+----------------------+
| a73a553e-6ea2-47dc-af84-6880f5b584b5 | Cirros               |
| a55aa773-050a-404e-9aca-a1519bb6ec65 | Debian Jessie 64-bit |
+--------------------------------------+----------------------+
neutron subnet-list
+--------------------------------------+----------------+----------------+---------------------------------------------------+
| id                                   | name           | cidr           | allocation_pools                                  |
+--------------------------------------+----------------+----------------+---------------------------------------------------+
| a642ba2b-d69c-4124-ad2c-95e3b935736f | private-subnet | 192.168.1.0/24 | {"start": "192.168.1.10", "end": "192.168.1.100"} |
| 23524d61-fc31-4354-81af-ed042301e1e3 | public-subnet  | 10.156.0.0/14  | {"start": "10.158.0.10", "end": "10.158.0.100"}   |
+--------------------------------------+----------------+----------------+---------------------------------------------------+

From an external network

In progress...

How it works

This installation method of Openstack on Grid'5000 offer to the user common tasks to :

• Reserve resources and manage a OAR job.
• Deploy nodes with Kadeploy.
• Install a Puppet server.
• Manage Puppet modules maintained by Openstack.

Rake

Rake is a simple ruby build program with capabilities similar to make. Tasks are defined in a Rakefile and can be listed with command :

rake -T

XP5K

XP5K is a small Ruby library to help Grid'5000 users to script their experiments using the Grid'5000 API :

• Submit one or several jobs, get the status, remove jobs, no needs to know the job ID.
• Create Roles (list of nodes dedicated for a specific role, ex : puppetserver, controller, computes), no needs to know the hostname of allocated resources.
• Create one or several deployment on nodes defined by a job, by roles.
• Extend the Rake DSL to :
  • create roles,
  • launch parallel SSH commands on a list of nodes or roles through a SSH gateway (the gateway allow you to launch experiments from your personal computer).
• Read an experiment configuration file xp.conf to customize your experiment instance.
• Manage a file .xp_cache containing the context of an experiment instance (jobs, roles, ...)

Main tasks

Main tasks are defined in the Rakefile and on tasks/*.rb files.

rake cmd                        # Launch command in parallel, need cmd=<command> and host=<role|FQDN>
rake grid5000:clean             # Clean all OAR jobs
rake grid5000:deploy            # Submit Kadeploy environment deployment
rake grid5000:jobs              # Submit OAR jobs
rake grid5000:status            # Get OAR jobs status
rake puppet:agent:install       # Install Puppet agent package on all nodes
rake puppet:agent:run           # Puppet Puppet agent on node host=<role|FQDN>
rake puppet:hiera:generate      # Generate hiera database
rake puppet:modules:get         # Download external openstack Puppet modules
rake puppet:modules:remove      # Delete external Puppet modules
rake puppet:modules:upload      # Upload Puppet modules and hiera database
rake puppet:server:bootstrap    # bootstrap Puppet server
rake run                        # Start Openstack deployment
rake shell                      # ssh on host, need host=<role|FQDN>

The run task is a flow of tasks that prepare the experiment and launch at the end the main task of a scenario scenario:main (explained in the section below).

Scenarios

Scenarios are placed in ./scenarios directory. The default scenario is liberty_starter_kit. To use another scenario, update the xp.conf file with :

scenario 'scenario_name'

A scenario must be at least composed by :

• A ./Puppetfile file, that contain a list of external Puppet modules used.
• A ./puppet/modules directory for Puppet modules dedicated to the scenario.
• A ./puppet/hiera directory, with :

├── generated (empty directory)
└── templates
    ├── common.yaml
    └── nodes
        └── puppetserver.yaml

common.yaml:

---
classes:
- xp
- xp::locales

puppetserver.yaml :

---
classes:
- xp::puppet::server
puppet::server::autosign:
- host1

• A ./tasks/scenario.rb file with at least :

# Scenario dedicated Rake task
#

# Define OAR job (required)
#
xp.define_job(@job_def)


# Define Kadeploy deployment (required)
#
xp.define_deployment(@deployment_def)


namespace :scenario do

  # Required task
  desc 'Main task called at the end of `run` task'
  task :main do

  end

end