policy

module
v1.0.2 Latest Latest
Warning

This package is not in the latest version of its module.

Go to latest
Published: Dec 15, 2022 License: Apache-2.0, BSD-3-Clause, MIT

README

pipeline status coverage report

Policy Service

The policy service provides HTTP API to evaluate/execute OPA policies.

It is developed using the Goa v3 framework and uses the Go OPA framework as a library.

While the service is up and running, you can see a live Swagger API description at servicehost:serviceport/swagger-ui. In the local docker-compose environment, the Swagger URL is available at http://localhost:8081/swagger-ui/

High-level Overview
flowchart LR
	A([client]) -- HTTP --> B[Policy API]
	subgraph policy
		B --> C[(policies DB)]
	end
	C --sync--- D[Git Server]
Policy Evaluation

The policy service exposes HTTP endpoints to evaluate/execute policies. The endpoint interface is conformant to the TSA requirements document.

To evaluate a policy a GET or POST request is sent to the evaluation URL. The example URL below is given for the local docker-compose environment. The host and port parts will be different for the different environments.

# URL with example policy group, name and version
http://localhost:8081/policy/gaiax/didresolve/1.0/evaluation

# URL with parameter placeholders
http://localhost:8081/policy/{group}/{policy}/{version}/evaluation

There are three parameters in the URL specifying which exact policy should be evaluated - group, policy and version. These parameters are also important during policy development (see below) as group and policy must be used as package name inside the policy source code file.

The body of the POST request can be empty, but if it's not empty, it must be JSON. It is passed directly to the policy execution runtime. Inside the policy it is accessed with the global variable name input. For example, if you pass to the evaluation endpoint the following JSON, it will be accessible by input.message:

{
  "message": "hello world"
}

Here is a complete example CURL request:

curl -X POST http://localhost:8081/policy/gaiax/didresolve/1.0/evaluation -d '{"message":"hello world"}'
Policy Locking

The service exposes HTTP endpoints to lock and unlock policies. Locking a policy means that it's not allowed for evaluation (execution). Unlocking a policy allows its evaluation/execution to proceed normally.

Lock a policy with POST request:

curl -X POST http://localhost:8081/policy/gaiax/didresolve/1.0/lock

Unlock a policy with DELETE request:

curl -X DELETE http://localhost:8081/policy/gaiax/didresolve/1.0/lock
Policy Storage

Policies (rego source code and metadata) are stored in a MongoDB collection policies, with one collection document representing one policy. A document contains additional policy state un-related to OPA and Rego, but necessary for implementing the TSA requirements (e.g. policy lock/unlock).

The database is used as read-only source of truth for the current policy state when policies need to be evaluated. Policy storage is updated externally from a separate component. The update process is automatically triggered by updating policy source code files in an external Git server.

flowchart LR
	A[Policy\nDeveloper] --git push/merge--> B[Git branch]
	subgraph Git Server
		B --> C[example/1.0/policy.rego]
		B --> D[example/2.0/policy.rego]
		B --> G[example/3.0/policy.rego]
	end
	C --> E[Sync]
	D --> E[Sync]
	G --> E[Sync]
	E --update--> F
	subgraph policy service
		F[(policies DB)]
	end
Policy Development

Policies are written in the Rego language. Please refer to the OPA documentation for detailed overview of Rego and OPA capabilities.

Some conventions must be followed when writing policies.

  1. The filename of the policy must follow rules for the naming and directory structure: the group, policy name and version are directories inside the Git repo and policy file must be named policy.rego. For example: /gaiax/example/1.0/policy.rego.
  2. In the same directory there could be a data file containing static JSON, which is automatically available for use during policy evaluation by using the data variable. The file must be named data.json. Example: /gaiax/example/1.0/data.json
  3. In the same directory there could be a configuration file containing information for getting static JSON data from external URL. The file must be named data-config.json. Example: /gaiax/example/1.0/data-config.json

Note that there should only be one of the two files data.json or data-config.json in the same directory. If both files exist in the same directory tha data from the data.json file will be eventually overwritten by the data acquired using the configuration from the data-config.json file. 4. The policy package name inside the policy source code file must exactly match the group and policy (name) of the policy.

What does it mean?

Let's see an example for the 1st convention.

package gaiax.example

allow {
    input.message == "hello world"
}

Next, the filename must be /gaiax/example/1.0/policy.rego. When such file is synchronized with the policy service (storage), the naming convention allows the service to understand which part is the policy group, which part is policy name and which part is version.

If we create the above policy and store it in the Git repo as /gaiax/example/1.0/policy.rego, after the Git server is synchronized with the Policy Storage, the policy service will automatically expose URLs for working with the policy at:

http://localhost:8081/policy/gaiax/example/1.0/evaluation
http://localhost:8081/policy/gaiax/example/1.0/lock

The 2nd rule for static data file naming is to make sure that file /gaiax/example/1.0/data.json is passed and is available to the evaluation runtime when a policy is evaluated at URL:

http://localhost:8081/policy/gaiax/example/1.0/evaluation

Static data is accessed within the Rego policy with data.someKey. Example: If the /gaiax/example/1.0/data.json file is:

{
  "name": "some name"
}

one could access the data using data.name within the Rego source code.

The 3rd rule for configuration file is to provide configurations for getting static JSON data from external URL. The file must contain a URL, an HTTP method and a period, after which an HTTP request is made to get the latest data.

The period must be added as duration e.g. 10h, 1h30m etc.

The file MAY contain body for the request. Example file contents:

{
  "url": "http://example.com/data.json?page=3",
  "method": "GET",
  "period": "10h",
  "body": {
    "key": "value"
  }
}

This means that every 10 hours an HTTP request is going to be made on the given URL, with GET method and the result is going to be stored as static data for this policy and passed during evaluation.

The 4th rule for package naming is needed so that a generic evaluation function can be mapped and used for evaluating all kinds of different policies. Without a package naming rule, there's no way the service can automatically generate HTTP endpoints for working with arbitrary dynamically uploaded policies.

Access HTTP Headers inside a policy

HTTP request headers are passed to the evaluation runtime on each request. They can be accessed through a built-in extension function named external.http.header(). It accepts as argument the name of the header in Canonical format. For example, inside Rego the value of a header named Authorization can be retrieved as follows:

package example.example

auth := external.http.header("Authorization")

Header names are passed to the Rego runtime in Canonical format. This means that the first character and any characters following a hyphen are uppercase and the rest are lowercase.

More examples, if the policy service receive a request with the following headers:

accept-encoding: gzip, deflate
Accept-Language: en-us
fOO: Bar
x-loCATion: Baz

Inside a policy these headers could be accessed as follows:

accept_encoding := external.http.header("Accept-Encoding")
accept_language := external.http.header("Accept-Language")
foo := external.http.header("Foo")
location := external.http.header("X-Location")
Policy Extensions Functions

A brief documentation for the available Rego extensions functions which can be used during policy development.

Policy Extensions Functions

You can also look at the source code in package regofunc to understand the inner-working and capabilities of the extension functions.

GDPR

GDPR

Dependencies

Dependencies

License

Apache 2.0 license

Directories

Path Synopsis
cmd
sync Module
nolint:revive
nolint:revive
gen
internal
regocache
Package regocache implements in-memory caching of policy data structure.
Package regocache implements in-memory caching of policy data structure.
regofunc
Package regofunc provides functions that extend the Rego runtime with additional capabilities and built-in functions which can be used when writing and evaluating Rego polices.
Package regofunc provides functions that extend the Rego runtime with additional capabilities and built-in functions which can be used when writing and evaluating Rego polices.
service/policy/policydata/policydatafakes
Code generated by counterfeiter.
Code generated by counterfeiter.
service/policy/policyfakes
Code generated by counterfeiter.
Code generated by counterfeiter.

Jump to

Keyboard shortcuts

? : This menu
/ : Search site
f or F : Jump to
y or Y : Canonical URL