Check out OPAL main repo here.
An OPAL custom fetch provider to bring authorization state from Postgres.
This fetcher is both:
- A fully functional fetch-provider for Postgres: can be used by OPAL to fetch data from Postgres DB.
- Serving as an example how to write custom fetch providers for OPAL and how to publish them as pip packages.
You can test this fetcher with the example docker compose file in this repository root. Clone this repo, cd
into the cloned repo, and then run:
docker compose up
this docker compose configuration already correctly configures OPAL to load the Postgres Fetch Provider, and correctly configures OPAL_DATA_CONFIG_SOURCES
to include an entry that uses this fetcher.
The official docker image only contains the built-in fetch providers. You need to create your own Dockerfile
(that is based on the official docker image), that includes this fetcher's pip package.
Your Dockerfile
should look like this:
FROM permitio/opal-client:latest
RUN pip install --no-cache-dir --user opal-fetcher-postgres
Say your special Dockerfile from step one is called custom_client.Dockerfile
.
You must build a customized OPAL container from this Dockerfile, like so:
docker build -t yourcompany/opal-client -f custom_client.Dockerfile .
Pass the following environment variable to the OPAL client docker container (comma-separated provider modules):
OPAL_FETCH_PROVIDER_MODULES=opal_common.fetcher.providers,opal_fetcher_postgres.provider
Notice that OPAL receives a list from where to search for fetch providers.
The list in our case includes the built-in providers (opal_common.fetcher.providers
) and our custom postgres provider.
Your DataSourceEntry objects (either in OPAL_DATA_CONFIG_SOURCES
or in dynamic updates sent via the OPAL publish API) can now include this fetcher's config.
Example value of OPAL_DATA_CONFIG_SOURCES
(formatted nicely, but in env var you should pack this to one-line and no-spaces):
{
"config": {
"entries": [
{
"url": "postgresql://postgres@example_db:5432/postgres",
"config": {
"fetcher": "PostgresFetchProvider",
"query": "SELECT * from city;",
"connection_params": {
"password": "postgres"
}
},
"topics": [
"policy_data"
],
"dst_path": "cities"
}
]
}
}
Notice how config
is an instance of PostgresFetcherConfig
(code is in opal_fetcher_postgres/provider.py
).
Values for this fetcher config:
- The
url
is actually a postgres dsn. You can set the postgres password in the dsn itself if you want. connection_params
are optional, if you want to include certain overrides outside the dsn.- Your
config
must include thefetcher
key to indicate to OPAL that you use a custom fetcher. - Your
config
must include thequery
key to indicate what query to run against postgres.
While trying to send requests to a Postgres data source, you may encounter that the request fails. This can be caused by the format of the config entry URL for which the standard is:
postgresql://<user>:<password>@<host>/<db>
It might be most common that this request fails due to the password field being incorrectly parsed by the underlying library called asyncpg
, which is one of the required libraries used within our OPAL custom data fetcher.
In order to solve the issue, you need to change the data source config entry URL to the format shown below:
postgresql://<host>/<db>?user=<user>&password=<password>
OPAL is an administration layer for Open Policy Agent (OPA), detecting changes to both policy and policy data in realtime and pushing live updates to your agents.
OPAL brings open-policy up to the speed needed by live applications. As your application state changes (whether it's via your APIs, DBs, git, S3 or 3rd-party SaaS services), OPAL will make sure your services are always in sync with the authorization data and policy they need (and only those they need).
Check out OPAL's main site at OPAL.ac.