Table of contents

• Description
• Installation
• Operation Instructions

Description

The FabFed is a Python library for a cross-testbed federation framework that (1) models the network experiment (or "slice") across the FABRIC testbed and federated testbeds and providers, and (2) provides workflow tools to stitch l2 and l3 networks between the testbeds and providers.

The FabFed code took the initial form from the Mobius API, and refactored and reinvented the slice modeling, user interface, data structure and stitching workflows.

The example below showcases network stitching across providers, a chi provider and a fabric provider. The configuration, while incomplete, highlights how fabfed-py expresses dependencies.

• For more details, refer to fabfed's workflow design
• For a complete example, refer to Fabric Chameleon Stitching

  1 resource:
  2
  3   - network:
  4       - chi_network:
  5             provider: '{{ chi.chi_provider }}'
  6             site: CHI@UC
  7
  8   - network:
  9       - fabric_network:
 10            provider: '{{ fabric.fabric_provider }}'
 11            site: 'STAR'
 12            stitch_with:
 13              - network: '{{ network.chi_network }}'

Installation

FabFed is available at PyPI.

pip install fabfed-py

Alternatively, you may install and test using the following commands:

pip install -e .
fabfed --help
fabfed stitch-policy --help
fabfed workflow --help
fabfed sessions --help

If using the CloudLab provider, the following portal-tools module is a required dependency:

pip install git+https://gitlab.flux.utah.edu/stoller/portal-tools.git

Operation Instructions

• Fabfed worflow configuration is specified across one or more .fab files. Fabfed does not care how these files are named. Fabfed simply loads all the .fab configuration files, assembles them and parses the assembled configuration.
• Fabfed will pickup any file ending with the .fab extension in the directory specified by the --config-dir. If this option is not present, the current directory is used.
• The --var-file option can be used to override the default value of any variable. It consists of a set of key-value pairs with each pair written as key: value. At runtime, all variables found in an assembled configuration must have a value other than None. The parser will halt and throw an exeption otherwise.
• The --session is a friendly name used to track a given workflow.
• Use the --help options shown above if in doubt.
• When stitching networks across provider use stitch-policy to discover available stitch information

# Example to view stitch policy from cloudlab to fabric
fabfed stitch-policy -providers "fabric,cloudlab"

# Validation
fabfed workflow --config-dir some_dir [--var-file some_var_file.yml] --session some_session -validate
fabfed workflow --config-dir some_dir [--var-file some_var_file.yml] --session some_session -stitch-info [-summary] [-json]

fabfed workflow --config-dir some_dir [--var-file some_var_file.yml] --session some_session -plan [-summary] [-json]

fabfed workflow --config-dir some_dir [--var-file some_var_file.yml] --session some_session -apply

fabfed workflow --config-dir some_dir [--var-file some_var_file.yml] --session some_session -show [-summary] [-json]

fabfed workflow --config-dir some_dir [--var-file some_var_file.yml] --session some_session -destroy

# Use this option to manage your workflow sessions
fabfed sessions -show