A task is the single unit of work. Example high level descriptions of tasks include:
For more information: Understanding Tasks
A Job, at it’s core, is merely an easy way to refer to a collection of tasks.
Jobs are uniquely identified by a job_id (UUID) and include the following information:
A template is a pre-configured job with a set of tasks that include the most common configurations for a given fuzz job. Templates are analogous to playbooks or recipes, entirely built on top of the SDK. The templates can be recreated as scripts calling the SDK or by executing the CLI.
As an example, the ‘libfuzzer basic’ template includes the following tasks:
At this time, templates are statically defined. In the future, OneFuzz will allow the owner of a OneFuzz instance to manage their own templates, allowing central management of how fuzzing tasks are defined.
Repro is short for ‘reproduction VMs’. These VMs are created on-demand to enable
debugging a crash in the same environment used for fuzzing over an SSH tunnel.
The repro VM creation automation includes downloading the task data related to
the crash, executing any setup scripts, and launching the application under test
within a debugger (cdb -server on Windows and gdbserver on Linux).
At this time, the automatic-debugger connect is functional only for file-based fuzzing targets (such as libfuzzer or AFL), however users can connect into VMs directly via SSH or RDP (Windows only) and have total control of the VM.
An Azure Blob Storage Container. Each fuzzing task has a set of required and potentially optional containers that are used in a specific context.
A pool is a group of nodes to run Tasks. Pools enable users to specify which group of nodes their task should run on.
Pools are defined by:
name: the Name of the poolos: The operating system of the node (linux or windows)
arch: The CPU architecture of the nodes (only x86_64 for now)managed: Is the pool made up of OneFuzz managed scalesetsA scaleset is an Azure Virtual Machine Scaleset, of which the entire lifecycle is managed by OneFuzz. All of the VMs in the scaleset are automatically setup to connect to the OneFuzz instance as a Node.
Scalesets can run on almost any Azure VM Image (use the URN for the image) or user-provided OS images.
A single compute host to run tasks. Right now, these are only VMs in a scaleset, though support for on-prem or third-party cloud-hosted nodes will be available in the future.