A JupiterOne integration ingests information such as configurations and other metadata about digital and physical assets belonging to an organization. The integration is responsible for connecting to data provider APIs and determining changes to make to the JupiterOne graph database to reflect the current state of assets. Managed integrations execute within the JupiterOne infrastructure and are deployed by the JupiterOne engineering team.
JupiterOne accounts may configure a number of instances of an integration, each
containing credentials and other information necessary for the integration to
connect to provider APIs. An integration is triggered by an event containing the
instance configuration. IntegrationInstance.config
is encrypted at rest and
decrypted before it is delivered to the integration execution handler.
Currently, the integration instance configuration user interface will need code changes to collect necessary information.
Local execution of the integration is started through execute.ts
(yarn start
), which may be changed to load development credentials into the
IntegrationInstance.config
. Use environment variables to avoid publishing
sensitive information to GitHub!
Integration projects must provide documentation for docs.jupiterone.io. This documentation should outline the credentials required by the data provider API (including specific permissions if the data provider allows scoping of credentials), which entities are ingested, and what relationships are created. At build time, this documentation will be placed in a docs folder inside dist so that it's included in the NPM module.
The documentation should be placed in docs/jupiterone-io
and named after the
package. For example, an AWS integration with the name "jupiter-integration-aws"
in package.json
should have its documentation in
docs/jupiterone-io/jupiter-integration-aws.md
. Any other files in
docs/jupiterone-io
will not be published. Also note that namespace is ignored,
so "jupiter-integration-aws" and "@jupiterone/jupiter-integration-aws" should
both name their docs file the same.
The first header in the documentation is used as the title of the document in the table of contents on docs.jupiterone.io, so it should be the name of the provider (E.G. "AWS").
The documentation is pushed to docs.jupiterone.io every time a new version of
the integration is specified in package.json
, so make sure it's up to date
every time you release a new version.
Integrations mutate the graph to reflect configurations and metadata from the provider. Developing an integration involves:
- Establishing a secure connection to a provider API
- Fetching provider data and converting it to entities and relationships
- Collecting the existing set of entities and relationships already in the graph
- Performing a diff to determine which entites/relationships to create/update/delete
- Delivering create/update/delete operations to the persister to update the graph
Run the integration to see what happens. You may use use Node to execute directly on your machine (NVM is recommended), or you may use Docker Compose to run the project in a contained environment
Node:
- Install Docker and Node
yarn install
- Provide credentials in
.env
yarn start:graph
yarn start
Docker Compose:
- Install Docker and Docker Compose
- Copy
docker-compose.sample.yml
todocker-compose.yml
- Provide credentials in
docker-compose.yml
ENV
docker-compose build
docker-compose run --rm integration yarn install
docker-compose run --rm integration yarn start
Activity is logged to the console indicating the operations produced and processed. View raw data in the graph database using Graphexp.
Execute the integration again to see that there are no change operations produced.
Restart the graph server to clear the data when you want to run the integration with no existing data.
yarn stop:graph && yarn start:graph
Or:
docker-compose restart graph
You can start interactive session inside a container:
docker-compose run --rm integration bash
Provider API configuration is specified by users when they install the integration into their JupiterOne environment. Some integrations may also require pre-shared secrets, used across all integration installations, which is to be secured by JupiterOne and provided in the execution context.
Local execution requires the same configuration parameters for a development
provider account. tools/execute.ts
is the place to provide the parameters. The
execution script must not include any credentials, and it is important to make
it easy for other developers to execute the integration against their own
development provider account.
- Update
tools/execute.ts
to provide the properties required by theexecutionHandler
function - Create a
.env
file to provide the environment variables transferred into the properties
For example, given this execution script:
const integrationConfig = {
apiToken: process.env.MYPROVIDER_LOCAL_EXECUTION_API_TOKEN,
};
const invocationArgs = {
preSharedPrivateKey: process.env.MYPROVIDER_LOCAL_EXECUTION_PRIVATE_KEY,
};
Create a .env
file (this is .gitignore
'd):
MYPROVIDER_LOCAL_EXECUTION_API_TOKEN=abc123
MYPROVIDER_LOCAL_EXECUTION_PRIVATE_KEY='something\nreally\nlong'
Environment variables can modify some aspects of the integration SDK behavior.
These may be added to your .env
with values to overrided the defaults listed
here.
GRAPH_DB_ENDPOINT
-"localhost"
All tests must be written using Jest. Focus on testing provider API interactions and conversion from provider data to entities and relationships.
To run tests locally:
yarn test
Or:
docker-compose run --rm integration yarn test
Managed integrations are deployed into the JupiterOne infrastructure by staff engineers using internal projects that declare a dependency on the open source integration NPM package. The package will be published by the JupiterOne team.