Skip to main content

Use Harness Policy As Code to enforce policies on steps

This topic explains steps to create policies using the Harness Policy As Code and apply them to your pipeline steps. Harness Policy As Code uses the Open Policy Agency (OPA) to store policies on the Harness Platform. For more information, go to Harness Policy As Code Overview.

Before you begin

Video of using policies with Harness steps

Sample pipeline

Here is a sample YAML of a pipeline with steps:

pipeline:
name: OPA Policy Pipeline
identifier: OPA_Policy_Pipeline
projectIdentifier: ## YOUR PROJECT ID
orgIdentifier: ## YOUR ORG ID
tags: {}
stages:
- stage:
name: OPA Sandbox
identifier: OPA_Sandbox
description: ""
type: Custom
spec:
execution:
steps:
- step:
type: ShellScript
name: Compute Variable
identifier: Compute_Variable
timeout: 10m
spec:
shell: Bash
onDelegate: true
source:
type: Inline
spec:
script: |-
export serviceMsg=<+pipeline.variables.msg>
echo serviceMsg
environmentVariables: []
outputVariables:
- name: serviceMsg
type: String
value: serviceMsg
enforce:
policySets:
- Variable_Sanitization
tags: {}
variables:
- name: msg
type: String
description: ""
value: <+input>

Set up your policy and policy set for the pipeline step

To apply a policy to a pipeline step, you must:

  1. Create a policy
  2. Create a policy set
  3. Add a policy set to a step in the pipeline
  4. Run the pipeline

Create a policy

The first step of using policies with your steps is navigating to policies and creating your own policy.

The policy will be coupled with how the step is designed and configured. If the step is more free-form like a shell script step, it's essential that you properly map and capture the outputs of variables you wish to reference in your policy step for validation.

More out-of-the-box steps like the rolling deployment step, have a pre-defined output variable path and make it easier for the policy to be re-used across multiple steps.

For this example, we will be modeling the policy around this shell script step

YAML snippet:

- step:
type: ShellScript
name: Compute Variable
identifier: Compute_Variable
timeout: 10m
spec:
shell: Bash
onDelegate: true
source:
type: Inline
spec:
script: |-
export serviceMsg=<+pipeline.variables.msg>
echo serviceMsg
environmentVariables: []
outputVariables:
- name: serviceMsg
type: String
value: serviceMsg

Create a policy with the following:

package pipeline

deny [msg] {
true
# Show a human friendly message
input[0].outcome.outputVariables["serviceMsg"]==""
msg:= "serviceMsg cannot be empty"

}

The policy above is checking to make sure that the message isn't empty. We want to enforce that all variables have some value that is not an empty value.

Create a policy set

Navigate to your Policy Set and click "+ New Policy Set".

  • Please provide a name for your policy - (i.e. Variable Sanitization)
  • Entity Type that this policy set applies to: Custom
  • On what event should the policy set be evaluated: On Step

Note: Entity type is custom because the step we are enforcing is a custom object type.

By setting the policy on step users' can now access the outputs and inputs of a step and enforce the policy on them.

Add a policy set to a step

You will need to add your policy set to your step in order for the policy engine to evaluate the outputs of the step and provide a judgment.

Navigate to the advanced section of the step and select the policy enforcement Accordion.

You will be prompted to pick a policy set to associate with the step.

After the selection, you will see the policy set associated with the step:

Run the pipeline

Don't provide any input into the Run form and just hit run, you will see the policy evaluate and error out with the below message:

You will see the policy evaluate and the corresponding denial message. If you re-run the pipeline and provide input, you will see the pipeline succeed.

In the example below, we provided deploy and the variable was assigned "deploy" and the validation passed

See also