Schema

In Infrahub, the schema is at the center of most things and our goal is to provide as much flexibility as possible to allow users to extend and customize the schema.

Out of the box, Infrahub doesn't have a schema for most things and it's up to users to load a schema that fits their needs. Over time we plan to maintain different schemas for the common types of use cases, but for now, we are providing one example schema to model a basic network with objects like Device, Interface, IPAddress, etc.

Unlike traditional databases that can only have one schema at a time, in Infrahub it is possible to have a different schema per branch. This is possible because the schema itself is stored in the database like any other object.

There are several ways to load a new schema.

note

To help with the development process of a schema definition file, you can leverage the schemas we publish for validation within your editor

Schema definition

Namespace, Node, Attributes, Relationships, and Generics

The schema is composed of 4 primary types of objects: Nodes- that are themselves composed of Attributes and Relationships and finally Generics.

A Node in Infrahub represents a Model.
An Attribute represents a direct value associated with a Node like a Text, a Number etc ...
A Relationship represents a link between 2 Node, a Relationship can be of cardinality one or many.
A Generic can be used to share some attributes between multiple Node, if you're familiar with programming concept, it's close to class inheritance.

In the example below, the node Person has 2 attributes (name and description) and the node Car has 1 attribute (model) and 1 relationship to Person, identified by owner.

nodes:
  - name: Person
    namespace: Example1
    attributes:
      - name: name
        kind: Text
        unique: true
      - name: description
        kind: Text
        optional: true
  - name: Car
    namespace: Example1
    attributes:
      - name: model
        kind: Text
    relationships:
      - name: owner
        peer: Example1Person
        optional: false
        cardinality: one
        kind: Attribute

Node, Attribute, and Relationship are defined by their kind. While the name and the namespace of the node are up to the creator of the schema, the kinds for the attributes and the relationships are coming from Infrahub. The kind of an attribute, or a relationship, is important because it defines how each element is represented in GraphQL and the UI.

info

The kind of a model is generated by concatenating the namespace and the name.

Attribute kinds

Text: Standard Text
Number: Standard Number
TextArea: Long-form Text that can span multiple lines
Boolean: Flag that can be either True or False
DateTime: A Data and a Time
Dropdown: A list of choices, each choice can have a color and description
Email: Email address
Password: A Text String that should be obfuscated
URL: An URL to a website or a resource over HTTP
File: Path to a file on the filesystem
MacAddress: Mac Address following the format (XX:XX:XX:XX:XX:XX)
Color: An HTML color
Bandwidth: Bandwidth in kbps
IPHost: IP Address in either IPV4 or IPv6 format
IPNetwork: IP Network in either IPV4 or IPv6 format
Checkbox: Duplicate of Boolean
List: List of any value
JSON: Any data structure compatible with JSON
Any: Can be anything

Relationship kinds

Generic: Default relationship without specific significance
Attribute: Relationship of type Attribute are represented in the detailed view and the list view
Component: Indicate a relationship with another node that is a component of the current node. Example: Interface is a component to a Device
Parent: Indicate a relationship with another node that is a parent to the current node. Example: Device is a parent to an Interface
Group: Indicate a relationship to a member or a subscriber of a group

Attribute kinds behavior in the UI

Kind	Display in List View	Display in Detailed View
`ID`	No	Yes
`Text`	Yes	Yes
`Number`	Yes	Yes
`Boolean`	Yes	Yes
`Dropdown`	Yes	Yes
`TextArea`	No	Yes
`DateTime`	No	Yes
`Email`	Yes	Yes
`Password`	No	Yes
`URL`	Yes	Yes
`File`	Yes	Yes
`MacAddress`	Yes	Yes
`Color`	Yes	Yes
`Bandwidth`	Yes	Yes
`IPHost`	Yes	Yes
`IPNetwork`	Yes	Yes
`Checkbox`	No	Yes
`List`	No	Yes
`JSON`	No	Yes
`Any`	No	Yes

Relationship kinds behavior in the UI

ID	cardinality	Display in List View	Display in Detailed View	Display in Tab
`Generic`	`one`	No	Yes	No
`Generic`	`many`	No	No	Yes
`Attribute`	`one`	Yes	Yes	No
`Attribute`	`many`	Yes	Yes	No
`Component`	`one`	No	Yes	No
`Component`	`many`	No	No	Yes
`Hierarchy`	`one`	Yes	Yes	No
`Hierarchy`	`many`	No	No	Yes
`Parent`	`one`	No	Yes	No
`Parent`	`many`	No	Yes	No

Uniqueness Constraints

More complex uniqueness constraints, composed of multiple attributes and/or relationships can be defined at the Node or Generic level with the property uniqueness_constraints. It's possible to define multiple uniqueness constraints and each of them will be evaluated individually.

In the example below, the node schema ExampleCar, ["owner", "model__value"] guarantee that a car will unique based on the owner and the model of the car.

uniqueness_constraints can be composed of a list of N number of attribute or relationship of cardinality one:

For an attribute, the valid format is <attributename>__value. Currently only value is supported but in the future the plan is to allow additional metadata to be used as well.
For a relationship, only the name of the relationship should be provided <relationshipname>. Only relationship of cardinality one are supported and the relationship must be mandatory.

nodes:
  - name: Person
    namespace: Example
    attributes:
      - name: name
        kind: Text
        unique: true
  - name: Car
    namespace: Example
    uniqueness_constraints:
      - ["owner", "model__value"]
    attributes:
      - name: model
        kind: Text
    relationships:
      - name: owner
        peer: ExamplePerson
        optional: false
        cardinality: one

Human friendly identifier (hfid)

In addition to the internal ID automatically generated for each object (UUID), Infrahub provide a way to define a set of fields that will constitute a human friendly identifier (HFID).

Use Case

The hfid is meant to be used to reference objects uniquely across systems, or even before an object has been created in Infrahub. Having a human friendly way to identify an object is very important to build robust data synchronization between systems and to provide true idempotent script.

In the network industry:

the hfid of a device could be its name : atl1-edge01
the hfid of an interface could be the combination of the name of the device and the name of the interface: (atl1-edge01, Ethernet1)

In the example below, each ExamplePerson, will have a hfid composed of his/her lastname and firstname.

human_friendly_id can be composed of N number of attribute or relationship of cardinality one:

For an attribute, the valid format is <attributename>__value. Currently only value is supported but in the future the plan is to allow additional metadata to be used as well.
For a relationship, the name of the relationship and the name of a unique attribute must be provided <relationshipname>__<attributename>__value. Only relationship of cardinality one are supported and the relationship must be mandatory.

nodes:
  - name: Person
    namespace: Example
    human_friendly_id: ["lastname__value", "firstname__value"]
    attributes:
      - name: firstname
        kind: Text
      - name: lastname
        kind: Text

info

If a human_friendly_id is defined on a node but no uniqueness_constraints are present, a uniqueness constraints matching the human_friendly_id will be automatically created.

If at least one attribute is marked as unique, and human_friendly_id hasn't been defined, human_friendly_id will be automatically created.

`hfid` support in GraphQL

hfid can be used as a alternative to id in most places in GraphQL, including :

Update Mutation
Upsert Mutation
Delete Mutation
Related Node in all Mutations

Generics

A generic can be used to:

Share multiple attributes or relationships between different types of nodes.
Connect multiple types of Node to the same relationship.
Define Attribute and Relationship on a specific list of nodes and avoid creating attributes for everything

In the example below, we took the schema that we used previously and refactored it using a generic Now Car is a generic with 2 attributes and 1 relationship and 2 models. ElectricCar and GazCar are referencing it. In the GraphQL schema, ElectricCar and GazCar will have all the attributes and the relationships of Car in addition to the one defined under their respective section.

generics:
  - name: Car
    namespace: Example1
    attributes:
      - kind: Text
        name: name
        unique: true
      - name: color
        kind: Colo
    relationships:
      - cardinality: one
        identifier: person__car
        name: owner
        optional: false
        peer: Example1Person
nodes:
  - name: ElectricCar
    namespace: Example1
    attributes:
      - kind: Number
        name: nbr_engine
    inherit_from: [ Example1Car ]
  - name: GazCar
    namespace: Example1
    attributes:
      - kind: Number
        name: mpg
    inherit_from: [ Example1Car ]
  - name: Person
    namespace: Example1
    attributes:
      - kind: Text
        name: name
        unique: true
      - kind: Number
        name: height
        optional: true
    relationships:
      - cardinality: many
        identifier: person__car
        name: cars
        peer: Example1Car

Inheritance between generics and nodes

Basic inheritance

Inheritance allows Nodes to derive properties from Generics. This relationship ensures that Nodes maintain a consistent set of attributes defined at the Generic level, reducing redundancy and simplifying updates.

Single generic inheritance

When a Node is derived from a single Generic, it inherits all properties defined in that Generic unless overridden at the Node level.

generics:
  - name: GenericDevice
    namespace: Infra
    description: "Generic Device object."
    label: "Device"
    icon: "mdi:server"
    default_filter: name__value
    order_by:
      - name__value
    display_labels:
      - name__value
    attributes:
      - name: name
        kind: Text
        unique: true
        order_weight: 1000
      - name: description
        kind: Text
        optional: true
        order_weight: 1100

nodes:
  - name: Device
    namespace: Infra
    inherit_from:
      - "GenericDevice"
    attributes:
      - name: role
        kind: Dropdown
        optional: true
        choices:
          - name: core
            label: Core Router
            description: Central part of the network.
            color: "#7f7fff"
          - name: spine
            label: Spine Router
            description: "Aggregation router part of a Fabric."
            color: "#aeeeee"
          - name: leaf
            label: Leaf Switch
            description: "Top of Rack part of a Fabric."
            color: "#e6e6fa"
        order_weight: 1400

Multiple generics inheritance

When multiple Generics are involved, determining which properties to inherit can be challenging. The implemented solution prioritizes the first Generic listed in the inherited_from array.

First Generic Preference

Properties will be inherited from the first Generic in the inherited_from list if not explicitly defined at the Node level. This approach simplifies conflict resolution.

generics:
  - name: GenericDevice
    namespace: Infra
    icon: "mdi:router"
    default_filter: "name__value"
    attributes:
      - name: name
        kind: Text
        unique: true
        order_weight: 1000

  - name: Asset
    namespace: Infra
    icon: "mdi:server"
    attributes:
      - name: asset_tag
        label: Asset Tag
        kind: Text
        unique: true
        order_weight: 1000

nodes:
  - name: Device
    namespace: Infra
    inherit_from:
      - "InfraGenericDevice"
      - "InfraAsset"

Inherited properties

List of inheritable properties

human_friendly_id: A user-friendly identifier for the entity.
display_labels: Labels used for display purposes.
default_filter: The default filter applied to the entity.
menu_placement: Defines where the entity appears in menus.
uniqueness_constraints: Constraints ensuring unique values for certain attributes.
icon: The icon representing the entity.
order_by: The default order in which entities are listed.

Limitations and Considerations

Inheritance of properties occurs only during the first import. If a Generic's properties are updated after the Node has been created, these updates will not be propagated to the Node. Users must manually update Nodes if they want to reflect changes made to Generics.

Branch support

By default, all models defined in the schema will be branch-aware which means that any changes to an object based on a branch-aware model will be local to the branch and will not affect the other branches.

A model can also be configured as:

branch agnostic: All changes to an object based on a branch agnostic model will automatically be available in all branches.
branch local: All changes will stay local to the branch. A model in branch local mode will not be affected by the Diff and the Merge.

Summary

Branch Support	Description	Diff	Merge	Rebase
Aware	All changes will be local to the branch and can be merged back into the main branch.	Yes	Yes	Yes
Agnostic	All changes will automatically be available in all branches	No	No	No
Local	All changes will be local to the branch and will not be merged to other branches.	No	No	Yes

Branch agnostic

In the frontend, the API, or the GraphQL endpoint branch-agnostic objects can be modified on any branch—no restrictions apply.

To configure a model as branch-agnostic you need to set the option branch to agnostic in the schema

nodes:
  - name: Fruit
    namespace: Example
    branch: "agnostic"
    attributes:
      - kind: Text
        name: name

Attributes and relationships

Attributes and relationships can be configured as branch-aware, branch-agnostic, or branch-local too, independently of the configuration of the model itself using the parameter branch.

nodes:
  - name: Fruit
    namespace: Example
    branch: "agnostic"
    attributes:
      - kind: Text
        name: name
        branch: "aware"

By default, if a specific value is not defined:

attributes will inherit the configuration of their parent model.
relationships will become:
- branch-agnostic only if both models on each end of the relationship are branch-agnostic. If either model is branch-aware the relationship will be set as branch-aware.
- branch-local if either model, on each end of the relationship, is branch-local.

Hierarchical mode

It's possible to organize some nodes of similar types in a hierarchy or a tree, to enable additional capabilities.

Examples

Groups can be organized in a hierarchy by default which makes it possible to query the members of all sub-groups at once.
Assuming we have defined a Person object connected to a City object, which itself is part of a hierarchy of Location (Region > Country > City) , it will be possible to query all Person per Country or per Region natively without having a direct relationship between Person and Country

A hierarchy must be defined around a specific Generic to ensure that all nodes that are part of a given hierarchy share some attributes. In the schema the attribute hierarchical: true on a Generic activate the hierarchical mode.

All nodes inheriting from this main Generic will automatically have the hierarchical mode enabled as well. Each node can only inherit from one Generic with hierarchical mode at the same time.

All nodes with hierarchical support will automatically have 2 new relationships:

parent of cardinality one
children of cardinality many

Specific the structure of the hierarchy

By default, any node inheriting from the main Generic can be a parent or a children, which might not be always desirable. It's possible to limit which type of models are valid parent or children in the schema with the attributes parent & children at the node level.

In the example below, we are defining a hierarchy of Location with Region, Country and City. Only Country can be defined as children of a Region and only City can be defined as children of a Country

version: '1.0'
generics:
  - name: Generic
    namespace: Location
    hierarchical: true
    attributes:
      - name: name
        kind: Text
nodes:
  - name: Region
    namespace: Location
    inherit_from: ["LocationGeneric"]
    parent: ""
    children: "LocationCountry"
  - name: Country
    namespace: Location
    inherit_from: ["LocationGeneric"]
    parent: "LocationRegion"
    children: "LocationCity"
  - name: City
    namespace: Location
    inherit_from: ["LocationGeneric"]
    parent: "LocationCountry"
    children: ""

Query data around a hierarchy in GraphQL

In GraphQL query , all nodes with hierarchical support will have 2 additional fields:

ancestors to access all ancestors of a given nodes (parent of parents`)
descendants to access all descendants of a given nodes (children of children)

info

ancestors and descendants are separated from parent and children to clearly indicate that we need to query the hierarchy and not the directly connected nodes.

For all relationships of cardinality many, it's possible to query all the objects associated with descendants nodes as well by providing the include_descendants flag.

As an example, The GraphQL query below will query the members of all groups that are defined as children of Group1. Assuming that Group1 has 2 sub-groups defined Group A and Group B (AKA children), the query will return the members of Group1, Group A and Group B together.

query {
    CoreStandardGroup(name__value: "Group1") {
        edges {
            node {
                id
                display_label
                members(include_descendants: true) {
                    count
                    edges {
                        node {
                            id
                            display_label
                        }
                    }
                }
            }
        }
    }
}

The position of a model in the frontend sidebar menu, is controlled by the schema itself. At the node (or generic) level, 3 attributes are available to control where a given model should be displayed in the menu and what icon it should have.

include_in_menu & menu_placement

include_in_menu is a Boolean to indicate if a given model should be displayed in the menu or not. By default, all models will be displayed directly under Objects in the frontend.

menu_placement makes it possible to create a nested structure within the menu to display a given model under another one.

Icon

icon can be used to indicate which icon should be displayed next to the model in the menu. Any icon from the Material Design Icons library / iconify.design can be used.

The full identifier of the icon must be used. Example mdi:domain

Examples

Schema
---
version: '1.0'
nodes:
  - name: Region
    namespace: Location
    icon: "mdi:web"
    include_in_menu: true
    label: "Region"
  - name: Site
    namespace: Location
    icon: "mdi:domain"
    include_in_menu: true
    menu_placement: "LocationRegion"
    label: "Site"
  - name: Rack
    namespace: Location
    icon: "mdi:server-outline"
    include_in_menu: true
    menu_placement: "LocationRegion"
    label: "Rack"

Rendering

Location Menu

Load a schema into Infrahub

Schema files can be loaded into Infrahub with the infrahubctl command or directly via the Git integration.

infrahubctl command

The infrahubctl command can be used to check & load individual schema files or multiple files as part of a directory.

The infrahub schema check command will validate if a given schema is valid and it will return a summary of the changes that will be applied to the schema if the schema was loaded.

infrahubctl schema check <path to schema file or a directory> <path to schema file or a directory> [--branch <branch_name>]

The infrahub schema load command will load the schemas into Infrahub into the specified branch.

infrahubctl schema load <path to schema file or a directory> <path to schema file or a directory> [--branch <branch_name>]

info

The infrahubctl schema load command can be used to both upload a new schema and update an existing one.

Git integration

You can defined a schema in an external repository. The schemas that should be loaded must be declared in the .infrahub.yml file, under schemas.

Individual files and directory are both supported.

---
schemas:
  - schemas/demo_edge_fabric.yml

Schema update and data migrations

Unlike most databases that support a single schema at a time, in Infrahub it is possible to have a different schema per branch. This is possible because the schema itself is stored in the database like any other object.

For every update to the schema, Infrahub will automatically execute some data validations and/or migrations to ensure that the data currently in the branch is compliant with the new schema.

The validations and the migrations will also be applied :

As part of a Proposed Change
During a branch rebase
During a branch merge

info

While a new version of the schema can be directly loaded into the main branch, the recommendation is to create a new branch and to integrate the changes into the main branch via a Proposed Change

State: absent or present

The format of the schema is declarative and incremental to allow schema to be composed from multiple sources.

To remove an existing Node, Generic, Attribute or Relationship from the schema it is required to use the state: absent flag to explicitly indicate that this element must be deleted. Without this flag the element will be ignored but it will not be removed.

  - name: Rack
    namespace: Infra
    state: absent
    attributes:
      - name: name
        kind: Text
    [..]

info

The default value is state: present and doesn't have to be provided on all elements of the schema.

Rename existing elements

All elements of the schema are identified by a human readable identifier that allow schema to be reused across environment:

For Node and Generic, the identifier is the Kind which is composed of namespace+name
For Attribute and Relationship the identifier is the name

Since namespace & name are used to identify the elements of the schema, if you need to update them after loading the schema an additional step will be required.

In order to update the human readable identifier, it's required to temporarily provide the internal identifier of the element directly in the configuration. All Node, Generic, Attribute or Relationship have their unique identifier that can be find easily in the schema explorer in the frontend http://localhost:8000/schema The internal identifier should be an UUID4, with 36 characters.

New Version
Previous Version

  - name: Rack
    namespace: Infra
    attributes:
      - id: 76df607c-883f-4cf6-9087-353dc2d863e5
        name: rack_name
        kind: Text
    [..]

  - name: Rack
    namespace: Infra
    attributes:
      - name: name
        kind: Text
    [..]

Restrictions and limitations

Most changes to an existing schema are supported but there are a few properties that can't be updated in the current release:

Properties that can't be updated on a Node or a Generic: branch
Properties that can't be updated on an Attribute: branch
Properties that can't be updated on an Relationship: branch, direction, hierarchical

A reference guide for validators and migrations is available

Known limitations

When updating the identifier on a relationship, the existing relationships won't be impacted or updated even if it was the last relationship using a given identifier. ISSUE-2472

Schema

Schema definition​

Namespace, Node, Attributes, Relationships, and Generics​

Attribute kinds​

Relationship kinds​

Uniqueness Constraints​

Human friendly identifier (hfid)​

hfid support in GraphQL​

Generics​

Inheritance between generics and nodes​

Basic inheritance​

Single generic inheritance​

Multiple generics inheritance​

Inherited properties​

List of inheritable properties​

Branch support​

Summary​

Branch agnostic​

Attributes and relationships​

Hierarchical mode​

Specific the structure of the hierarchy​

Query data around a hierarchy in GraphQL​

Menu​

include_in_menu & menu_placement​

Icon​

Examples​

Rendering​

Load a schema into Infrahub​

infrahubctl command​

Git integration​

Schema update and data migrations​

State: absent or present​

Rename existing elements​

Restrictions and limitations​

Known limitations​

Schema definition

Namespace, Node, Attributes, Relationships, and Generics

Attribute kinds

Relationship kinds

Uniqueness Constraints

Human friendly identifier (hfid)

`hfid` support in GraphQL

Generics

Inheritance between generics and nodes

Basic inheritance

Single generic inheritance

Multiple generics inheritance

Inherited properties

List of inheritable properties

Branch support

Summary

Branch agnostic

Attributes and relationships

Hierarchical mode

Specific the structure of the hierarchy

Query data around a hierarchy in GraphQL

Menu

include_in_menu & menu_placement

Icon

Examples

Rendering

Load a schema into Infrahub

infrahubctl command

Git integration

Schema update and data migrations

State: absent or present

Rename existing elements

Restrictions and limitations

Known limitations