SHIELD Documentation

SHIELD Architecture

SHIELD Architectural Overview

Feel free to skim this section on a first reading. It is definitely worth a thorough review, however, especially given how pervasive these ideas are in SHIELD.


The SHIELD Core is the heart, mind, and soul of a SHIELD installation. It is responsible for scheduling, metadata, configuration, monitoring, and much more.

The core is composed of several discrete components that all cooperate to deliver on the promise of data protection. They include:

  1. The SHIELD Database
  2. The SHIELD Vault
  3. The SHIELD Scheduler
  4. SHIELD Workers

These are described below, in more detail.


Without SHIELD Agents, nothing would get done. These distributed bits of software are reponsible for executing all manner of tasks, from backups, to restores, to archive purgation and validation testing.

Each agent uses locally installed plugins to interact with targets and stores to run backups, perform restores, prune archives, etc.


Plugins enable SHIELD to communicate with several different target data systems, and cloud storage solutions. These standalone executables operate according to a known protocol,

and allow operators and system integrators to support data systems and storage backends we haven't even dreamt up yet.

For a list of the currently implemented plugins that ship with SHIELD itself, see our plugin documentation.


A target is a data system, and they vary wildly. A PostgreSQL database can be a target. So can a Consul key-value store, or a Redis installation. If there's data (or live configuration that you don't want to lose), you can bet SHIELD thinks of it as a target.

SHIELD Agents interact with targets via target plugins.


A store is a storage system, usually off-site, redundant, remote, or all three. SHIELD stores archives in these systems, and then retrieves those archives for restoration tasks, later. Prominent, well-known stores include Amazon S3, WebDAV, Backblaze, etc.

SHIELD Agents interact with stores via store plugins.


At the heart of the SHIELD Core is the scheduler. It is responsible for regularly checking all defined jobs to see if they need to be executed (as tasks). It also handles a series of other tasks, including storage tests, archive pruning, etc.


A job ties together a target data system, a cloud store, and supplies scheduling and archive retention configuration. When scheduled, a job becomes a task.


The smallest schedulable unit in SHIELD, a task represents some specific operation, with configuration values, that will be executed at most once. A job turns into a task when it is scheduled by the SHIELD Core.

SHIELD also uses tasks for other purposes; pruning archives that have outlived their retention policy is handled via prune tasks. Cloud stores are validated regularly via test-store tasks.


A backup archive is the output of a successful backup task, and contains the encrypted and optionally compressed data that was extracted from the target data system. Archives are kept in cloud stores until they outlive their expiration, as set by the job that effected the task.


Internally, the SHIELD Core runs a set of worker threads that manage all communication with SHIELD Agents. These workers handle the execution of tasks on specific agents, stream logs back into the core, update task status as execution proceeds, and more.


The SHIELD Vault is a secure credentials storage solution. SHIELD uses the vault to store sensitive information like encryption parameters.

Web Interface

The Web Interface (or just web UI) is a graphical, web-based management interface for interacting with SHIELD. It provides modest monitoring capabilities, and allows operators to configure their SHIELD installations.


The SHIELD CLI is a command-line interface to SHIELD. It allows operators who do not wish to use a graphical interface to manager their SHIELD installations.

The CLI can also be put to use in a variety of automation context, as it is built to be scripted (as long as you can grok JSON output).


SHIELD features a rich and robust REST API, using JSON as a baseline payload format for both requests and responses. All approved external management utilities, including the Web UI and the CLI, exclusively use this API to do their jobs, ensuring that the API is featureful and complete.


The SHIELD Database is where non-sensitive, persistent data is stored. This includes everything from target configuration, to job schedules, to task histories and logs.

SHIELD uses the excellent SQLite3 embedded database system.