Skip to content


The Classifier System allows an operator to flexibly define groups of test/action sequences to perform on a Machine. These can be used to do any number of tasks. Typically the classifier allows an operator to add additional information to a Machine object that will alter the behavior of subsequent Workflow stages/tasks.

By default the classifier provides a limited set of tests and actions that can be performed. However, the operator can specify additional groups of test/actions pairs by use of the custom functions feature. See the Param classify/custom-functions documentation on how to use it.

By default, the classifier is designed to exit on the first test match that succeeds with a "pass" status. The operator can override this behavior by setting the optional continue value set to true in the classify/classification-data structure.

If you are extending the Classifier to add custom groups of test/action sequences, please review the function ... {} BASH functions in the template for examples.


The Classifier can operate in two modes, defined by the classify/version parameter. The main difference between these modes is that version 2 uses params that specify the names of other params. Those indirectly referenced params provide the classification data either through configuration or default value.

Version 2 is the preferred way of using the classifier, but classify/version defaults to 1 for historical purposes. Both are driven from the same classify task.

Classifier 1

Classifier version 1 is configured on a profile, usually the global profile. It has some limitations if you're trying to do pipelines or need to be able to have multiple sets of classification rules.

It uses the following params:

  • classify/classification-data: list of tests and actions for each test
  • classify/custom-functions: list of templates with the custom functions
  • classify/disable-classifier: default true, disables classifier

Classifier 2

Classifier version 2 is intended to be configured on a stage. This allows for creating stages that do their own classification rather than being limited to one classifier configuration for the whole workflow.

A version 2 classifier stage uses the following params:

  • classify/data-parameter: name of param that has classification-data
  • classify/function-parameter: name of param that has the list of templates
  • classify/disable-parameter: name of param that disables classifier

On stages with the following configuration, stages in the stage-list-parameter param will automatically be inserted into the workflow.

  • Tasks: [classify-stage-list-start, classify-stage-list-stop]
  • Params.classify/stage-list-parameter: name of param that has a list of stages

One of the stages in the stage list parameter is usually a version 2 classifier stage.


The parameters used by the version 2 classifier usually have default values specifically so users can inject their own modifications via params or profiles without having to make changes to the original content pack.

Getting Started with the Classifier

The Classifier rulesets of test/action sequences are configured by setting the Param classify/classification-data (v1) or being the target of a classify/data-parameter (v2) on a Machine. The usual rules and orders of precedence are followed on this Param (Param on Machine; Profile on Machine; Global Profile; etc...).

The classification data structure is a list of YAML or JSON objects, with two required and one optional set of values. The structure is as follows (example in YAML):

  - test: "<TEST_TYPE>"
      - "<ACTION>"
      - "<ACTION>"
    continue: true

The test and actions are required elements. continue is optional based on the operators need. continue is true by default.

Here are some example usage scenarious to help you get started.

Simple MAC Address Match and Classify

In this scenario, you will use the MAC Address as a simple test matcher to apply actions. Demonstrated in this example are adding two Params, and two profiles to the machine.

The Params in this example sets hostname, and universal/application value to specify the role that this machine will perform. Subsequent content/plugins can react accordingly to this. The profiles might be responsible for carrying configuration data for subsequent content/plugins to operate on. For example, setting BIOS bios-target-configuraion, determining what Firmware/Flash versions to use, Operating System configuration profile data (eg VMware ESXi configuration, etc.).

Last, this classify test/action sequence will continue on in the list to any subsequent test/action sequences after a match is found (continue: true).

  - test: "has_mac 24:6e:96:6a:40:34"
      - "set_parameter hostname r3r6-u03"
      - "set_parameter universal/application vcf381"
      - "add_profile demo-r3r6-machine-r3r6-u03"
      - "add_profile demo-r3r6-application-vcf381"
    continue: true

Change Workflow

In this scenario, the Classify system will be used to change the Machines Workflow. We will match on the Inventory field inventory/Manufacturer with a match of QEMU (which is typically how a KVM hypervisor will present a KVM backed Virtual Machine). The Machine will be changed to the centos-install workflow to install CentOS operating system.


When the Machine's workflow is changed, any remaining Stages/Tasks from the current workflow will be immediately replaced by the new Workflow.

This example assumes that the operator has already set the classify/custom-functions to include the template included already.

  - test: "has_inventory_value Manufacturer QEMU"
      - "change_workflow centos-install"

Since there is no continue: true setting, the classifier will exit after this rule, and will not process any further rules listed after this one.


The Flexiflow content pack is designed to provide controlled ways to inject new tasks and switch workflows. You may wish to utilize the constructs in that content pack over the change_workflow action in the Classifier.


The RackN Portal parses special annotations from templates referenced by the custom function params. These annotations must be prefixed with two hash symbols (##).

Name Example Description
@type @type name ^regex$ Allows for the definition of custom types to be used in @action or @test.
none Description A comment without an annotation next to the following annotations will be considered the function description.
@name @name Display Name Changes the name displayed in the portal for the test or action.
@action @action funcname(argName: type, arg2: type) For a classifier action, describes the shell function name, argument names and types.
@test @test funcname(argName: type, arg2: type) For a classifier test, describes the shell function name, argument names and types.
@hint @hint argName Hint text Provides a hint for what the named argument in a function does.
@icon @icon iconname Sets an icon for this function in the RackN Portal.
@deprecated @deprecated Message Displays a deprecation warning when this function is used in the RackN Portal.

The @type annotation uses JavaScript Regular Expressions. It does not have to be next to the other annotations and can be shared between multiple functions. If no type annotation is found for the type on a function, the type is ignored. If a type annotation is found in the RackN Portal and the user input value does not match the pattern, the input is colored red.

Annotation Example

## @type yesno ^(yes|no)$

## The description is not in any annotation
## @name Name of the action displayed in the Portal
## @action my_action(arg1: yesno, arg2: string)
## @hint arg1 enter yes or no
## @hint arg2 custom placeholder text
## @icon pencil
## @deprecated Deprecation warning text
function my_action() {
    echo "arg 1 $1"
    echo "arg 2 $2"
    exit 0

## Passes if the argument is yes
## @name Is Yes
## @test cl_is_yes(arg: yesno)
## @hint arg yes or no
## @icon check
function cl_is_yes() {
  if [[ "$1" == "yes" ]]; then
    echo "pass"
    echo "fail"