OPA can periodically download bundles of policy and data from remote HTTP servers. The policies and data are loaded on the fly without requiring a restart of OPA. Once the policies and data have been loaded, they are enforced immediately. Policies and data loaded from bundles are accessible via the standard OPA REST API.

Bundles provide an alternative to pushing policies into OPA via the REST APIs. By configuring OPA to download bundles from a remote HTTP server, you can ensure that OPA has an up-to-date copy of policies and data required for enforcement at all times.

By default, the OPA REST APIs will prevent you from modifying policy and data loaded via bundles. If you need to load policy and data from multiple sources, see the section below.

See the Configuration Reference for configuration details.

Bundle Service API

OPA expects the service to expose an API endpoint that serves bundles. The bundle API should allow clients to download bundles at an arbitrary URL. In combination with a service’s url path.

  1. GET /<service path>/<resource> HTTP/1.1

If the bundle exists, the server should respond with an HTTP 200 OK status followed by a gzipped tarball in the message body.

  1. HTTP/1.1 200 OK
  2. Content-Type: application/gzip

Enable bundle downloading via configuration. For example:

  1. services:
  2. - name: acmecorp
  3. url: https://example.com/service/v1
  4. credentials:
  5. bearer:
  6. token: "bGFza2RqZmxha3NkamZsa2Fqc2Rsa2ZqYWtsc2RqZmtramRmYWxkc2tm"
  7. bundles:
  8. authz:
  9. service: acmecorp
  10. resource: somedir/bundle.tar.gz
  11. persist: true
  12. polling:
  13. min_delay_seconds: 10
  14. max_delay_seconds: 20
  15. signing:
  16. keyid: my_global_key
  17. scope: read

Using this configuration, OPA will fetch bundles from https://example.com/service/v1/somedir/bundle.tar.gz.

The URL is constructed as follows:

  1. https://example.com/service/v1/somedir/bundle.tar.gz
  2. ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ ^^^^^^^^^^^^^^^^^^^^^^^^^^^
  3. services[0].url resource

If the bundles[_].resource field is not defined, the value defaults to bundles/<name> where the name is the key value in the configuration. For the example above this is authz and would default to bundles/authz.

Bundle names can have any valid YAML characters in them, including /. This can be useful when relying on default resource behavior with a name like authz/bundle.tar.gz which results in a resource of bundles/authz/bundle.tar.gz.

OPA can optionally persist activated bundles to disk for recovery purposes. To enable persistence, set the bundles[_].persist field to true. When bundle persistence is enabled, OPA will attempt to read the bundle from disk on startup. This allows OPA to start with the most recently activated bundle in case OPA cannot communicate with the bundle server. When communication between OPA and the bundle server is restored, the latest bundle is downloaded, activated, and persisted.

By default, bundles are persisted under the current working directory of the OPA process (e.g., ./.opa/bundles/<bundle-name>/bundle.tar.gz).

The optional bundles[_].signing field can be used to specify the keyid and scope that should be used for verifying the signature of the bundle. See this section for details.

See the following section for details on the bundle file format.

Note: The bundle config keyword will still work with the current versions of OPA, but has been deprecated. It is highly recommended to switch to the bundles configuration.

Caching

Services implementing the Bundle Service API should set the HTTP Etag header in bundle responses to identify the revision of the bundle. OPA will include the Etag value in the If-None-Match header of bundle requests. Services can check the If-None-Match header and reply with HTTP 304 Not Modified if the bundle has not changed since the last update.

Bundle File Format

Bundle files are gzipped tarballs that contain policies and data. The data files in the bundle must be organized hierarchically into directories inside the tarball.

The hierarchical organization indicates to OPA where to load the data files into the the data Document.

You can list the content of a bundle with tar.

  1. $ tar tzf bundle.tar.gz
  2. .manifest
  3. roles
  4. roles/bindings
  5. roles/bindings/data.json
  6. roles/permissions
  7. roles/permissions/data.json
  8. http
  9. http/example
  10. http/example/authz
  11. http/example/authz/authz.rego

In this example, the bundle contains one policy file (authz.rego) and two data files (roles/bindings/data.json and roles/permissions/data.json). The bundle may also contain an optional wasm binary file (policy.wasm). It stores the WebAssembly compiled version of all the Rego policy files within the bundle.

Bundle files may contain an optional .manifest file that stores bundle metadata. The file should contain a JSON serialized object, with the following fields:

  • revision - If the bundle service is capable of serving different revisions of the same bundle, the service should include a top-level revision field containing a string value that identifies the bundle revision.

  • roots - If you expect to load additional data into OPA from outside the bundle (e.g., via OPA’s HTTP API) you should include a top-level roots field containing of path prefixes that declare the scope of the bundle. See the section below on managing data from multiple sources. If the roots field is not included in the manifest it defaults to [""] which means that ALL data and policy must come from the bundle.

  • wasm - A list of OPA WebAssembly (Wasm) module files in the bundle along with metadata for how they should be evaluated. The following keys are supported:

    • entrypoint - A string path defining what query path the wasm module is built to evaluate. Once loaded any usage of this path in a query will use the Wasm module to compute the value.
    • module - A string path to the Wasm module relative to the root of the bundle.
  • metadata - An optional key that contains arbitrary metadata to accompany the bundle. This metadata is available for querying using data.system, along with the rest of the manifest.

For example, this manifest specifies a revision (which happens to be a Git commit hash) and a set of roots for the bundle contents. In this case, the manifest declares that it owns the roots data.roles and data.http.example.authz.

  1. {
  2. "revision" : "7864d60dd78d748dbce54b569e939f5b0dc07486",
  3. "roots": ["roles", "http/example/authz"]
  4. }

Another example, this time showing a Wasm module configured for data.http.example.authz.allow:

  1. {
  2. "revision": "7864d60dd78d748dbce54b569e939f5b0dc07486",
  3. "roots": ["roles", "http/example/authz"],
  4. "wasm": [
  5. {
  6. "entrypoint": "http/example/authz/allow",
  7. "module": "path/to/policy.wasm"
  8. }
  9. ]
  10. }

Some important details for bundle files:

  • OPA will only load data files named data.json or data.yaml (which contain JSON or YAML respectively). Other JSON and YAML files will be ignored.

  • The *.rego policy files must be valid Modules

  • OPA will only load Wasm modules named policy.wasm. Other WebAssembly binary files will be ignored.

YAML data loaded into OPA is converted to JSON. Since JSON is a subset of YAML, you are not allowed to use binary or null keys in objects and boolean and number keys are converted to strings. Also, YAML !!binary tags are not supported.

Multiple Sources of Policy and Data

By default, when OPA is configured to download policy and data from a bundle service, the entire content of OPA’s policy and data cache is defined by the bundle. However, if you need to load OPA with policy and data from multiple sources, you can implement your bundle service to generate bundles that are scoped to a subset of OPA’s policy and data cache.

🚨 We recommend that whenever possible, you implement policy and data aggregation centrally, however, in some cases that’s not possible (e.g., due to latency requirements.). When using multiple sources there are no ordering guarantees for which bundle loads first and takes over some root. If multiple bundles conflict, but are loaded at different times, OPA may go into an error state. It is highly recommended to use the health check and include bundle state: Monitoring OPA

To scope bundles to a subset of OPA’s policy and data cache, include a top-level roots key in the bundle that defines the roots of the data namespace that are owned by the bundle.

For example, the following manifest would declare two roots (acmecorp/policy and acmecorp/oncall):

  1. {
  2. "roots": ["acmecorp/policy", "acmecorp/oncall"]
  3. }

If OPA was loaded with a bundle containing this manifest it would only erase and overwrite policy and data under these roots. Policy and data loaded under other roots is left intact.

When OPA loads scoped bundles, it validates that:

  • The roots are not overlapping (e.g., a/b/c and a/b are overlapped and will result in an error.) Note: This is not enforced across multiple bundles. Only within the same bundle manifest.

  • The policies in the bundle are contained under the roots. This is determined by inspecting the package statement in each of the policy files. For example, given the manifest above, it would be an error to include a policy file containing package acmecorp.other because acmecorp.other is not contained in either of the roots.

  • The data in the bundle is contained under the roots.

If bundle validation fails, OPA will report the validation error via the Status API.

Debugging Your Bundles

When you run OPA, you can provide bundle files over the command line. This allows you to manually check that your bundles include all of the files that you intended and that they are structured correctly. For example:

  1. opa run bundle.tar.gz

Signing

To ensure the integrity of policies (ie. the policies are coming from a trusted source), policy bundles may be digitally signed so that industry-standard cryptographic primitives can verify their authenticity.

OPA supports digital signatures for policy bundles. Specifically, a signed bundle is a normal OPA bundle that includes a file named .signatures.json that dictates which files should be included in the bundle, what their SHA hashes are, and of course is cryptographically secure.

When OPA receives a new bundle, it checks that it has been properly signed using a (public) key that OPA has been configured with out-of-band. Only if that verification succeeds does OPA activate the new bundle; otherwise, OPA continues using its existing bundle and reports an activation failure via the status API and error logging.

⚠️ opa run performs bundle signature verification only when the -b/--bundle flag is given or when Bundle downloading is enabled. Sub-commands primarily used in development and debug environments (such as opa eval, opa test, etc.) DO NOT verify bundle signatures at this point in time.

Signature Format

Recall that a policy bundle is a gzipped tarball that contains policies and data. A signed bundle differs from a normal bundle in that it has a .signatures.json file as well.

  1. $ tar tzf bundle.tar.gz
  2. .manifest
  3. .signatures.json
  4. roles
  5. roles/bindings
  6. roles/bindings/data.json

The signatures file is a JSON file with an array of JSON Web Tokens (JWTs) that encapsulate the signatures for the bundle. Currently, you will be limited to one signature, as shown below. In the future, we may add support to include multiple signatures to sign different files within the bundle.

  1. {
  2. "signatures": [ "eyJhbGciOiJIUzI1NiIsInR5cCI6IkpXVCJ9.eyJmaWxlcyI6W3sibmFtZSI6Ii5tYW5pZmVzdCIsImhhc2giOiJjMjEzMTU0NGM3MTZhMjVhNWUzMWY1MDQzMDBmNTI0MGU4MjM1Y2FkYjlhNTdmMGJkMWI2ZjRiZDc0YjI2NjEyIiwiYWxnb3JpdGhtIjoiU0hBMjU2In0seyJuYW1lIjoicm9sZXMvYmluZGluZ3MvZGF0YS5qc29uIiwiaGFzaCI6IjQyY2ZlNjc2OGI1N2JiNWY3NTAzYzE2NWMyOGRkMDdhYzViODEzNTU0ZWJjODUwZjJjYzM1ODQzZTcxMzdiMWQifV0sImlhdCI6MTU5MjI0ODAyNywiaXNzIjoiSldUU2VydmljZSIsImtleWlkIjoibXlQdWJsaWNLZXkiLCJzY29wZSI6IndyaXRlIn0.ZjtUgXC6USwmhv4XP9gFH6MzZwpZrGpAL_2sTK1P-mg"]
  3. }

The JWT has the standard headers alg (for algorithm), typ (always JWT), and kid (for key id). It has a JSON payload of the following form:

  1. {
  2. "files": [
  3. {
  4. "name": ".manifest",
  5. "hash": "c2131544c716a25a5e31f504300f5240e8235cadb9a57f0bd1b6f4bd74b26612",
  6. "algorithm": "SHA-256"
  7. },
  8. {
  9. "name": "roles/bindings/data.json",
  10. "hash": "42cfe6768b57bb5f7503c165c28dd07ac5b813554ebc850f2cc35843e7137b1d"
  11. }
  12. ],
  13. "iat": 1592248027,
  14. "iss": "JWTService",
  15. "scope": "write"
  16. }
FieldTypeRequiredDescription
files[].namestringYesPath of a file in the bundle.
files[].hashstringYesOutput of the hashing algorithm applied to the file.
files[_].algorithmstringYesName of the hashing algorithm.
scopestringNoRepresents the fragment of signings.
iatstringNoTime of signature creation since epoch in seconds. For informational purposes only.
issstringNoIdentifies the issuer of the JWT. For informational purposes only.

Note: OPA will first look for the keyid on the command-line. If the keyid is empty, OPA will look for it in it’s configuration. If keyid is still empty, OPA will finally look for kid in the JWT header.

The following hashing algorithms are supported:

  1. MD5
  2. SHA-1
  3. SHA-224
  4. SHA-256
  5. SHA-384
  6. SHA-512
  7. SHA-512-224
  8. SHA-512-256

To calculate the digest for unstructured files (ie. all files except JSON or YAML files), apply the hash function to the byte stream of the file.

For structured files, read the byte stream and parse into a JSON structure; then recursively order the fields of all objects alphabetically and then apply the hash function to the result to compute the hash. This ensures that the digital signature is independent of whitespace and other non-semantic JSON features.

To generate a .signatures.json file for policy and data files that will be part of a bundle, see the opa sign command.

Signature Verification

When OPA receives a policy bundle that doesn’t include the .signatures.json file and the bundle is not configured to use a signature, OPA does not perform signature verification and activates the bundle just as it always has.

If the actual bundle contains the .signatures.json file but the bundle is not configured to use a signature, verification fails.

.signatures.json existsbundle configured to verify signatureverification performedresult
nononoNA
noyesyesfail
yesnoyesfail
yesyesyesdepends on the verification steps described below

When OPA receives a signed bundle it opens the .signatures.json file, grabs the JWT and performs the following steps:

  • Verify the JWT signature with the appropriate public key

  • Verify that the JWT payload and target directory specify the same set of files

  • Verify the content of each file by checking the hash recorded in the JWT payload is the same as the hash generated for that file

OPA activates the new bundle only if all the verification steps succeed; otherwise, it continues using its existing bundle and reports an activation failure via the status API and error logging.

The signature verification process uses each of the fields in the JWT header and payload as follows:

  • files: This list of files in the payload must match exactly the files in the bundle, and for each file the hash of the file must match

  • kid: If supplied in the header, dictates which key (and algorithm) to use for verification. The actual key is supplied via OPA out-of-band

  • scope: If supplied in the payload, must match exactly the value provided out-of-band to OPA

  • iat: unused for verification even if present in payload

  • iss: unused for verification even if present in payload

Signature Plugin

OPA supports the option to implement your own bundle signing and verification logic. This will be unnecessary for most and is intended for advanced use cases, such as leveraging key-related services from cloud providers. To implement your own signing and verification logic, you’ll need to extend OPA. Here is an example to get you started.

When registering custom signing and verification plugins, you will need to register the Signer and the Verifier under the same plugin key, because the plugin key is stored in the signed bundle and informs OPA which Verifier is capable of verifying the bundle, e.g.

  1. bundle.RegisterSigner("custom", &CustomSigner{})
  2. bundle.RegisterVerifier("custom", &CustomVerifier{})

Implementations

The Bundle API is simple. Most HTTP servers capable of serving static files will do. While not strictly required in all deployments, it is also good if the implementation supports:

  • HTTP caching using the ETag header. This keeps OPA from having to download a bundle unless the bundle’s content have changes.
  • Authentication. When exposing a bundle at a remote endpoint, it is often desirable to protect the data by requiring all requests to the endpoint to be authenticated.

This document lists some of the more common HTTP servers suitable as bundle servers, along with instructions for how to set them up as such.

Amazon S3

OPA Bundle Support

FeatureSupported
Caching headersYes
Authentication methodsAWS Signature

Setup Instructions

  1. Search for “S3” and on the “Buckets” page, click “Create bucket”.
  2. Fill in the form according to your preferences (name, region, etc).
  3. Either choose “Block all public access” for internal systems, or unmark the checkbox for that to allow external (authenticated) requests.
  4. You can now upload your bundle to the bucket. If you try to download it right away you’ll notice that by default you’re unauthorized to do so.
  5. To allow anyone to read the bundle, click on it and select “Make public” from the “Object actions” dropdown menu. If not, proceed to configure authentication.

Authentication

Authentication can be configured to either use the credentials of a service account stored in the environment, or to use credentials fetched from the AWS metadata API. The latter is only available from services running inside of AWS (on EC2 or ECS).

Both methods are going to need a policy for either the service account or the IAM role, so when that is mentioned in the steps for either method you may refer to the example below.

Example IAM policy

  1. {
  2. "Version": "2012-10-17",
  3. "Statement": [
  4. {
  5. "Effect": "Allow",
  6. "Action": [
  7. "s3:ListBucket"
  8. ],
  9. "Resource": [
  10. "arn:aws:s3:::my-example-opa-bucket"
  11. ]
  12. },
  13. {
  14. "Effect": "Allow",
  15. "Action": [
  16. "s3:PutObject",
  17. "s3:GetObject"
  18. ],
  19. "Resource": [
  20. "arn:aws:s3:::my-example-opa-bucket/*"
  21. ]
  22. }
  23. ]
  24. }

NOTE: The above policy permits both uploads and downloads, which is good for testing. The OPA client however needs only the s3:GetObject permission for downloads and should be the only permission granted for production use cases.

Environment Credentials
  1. Go to the “IAM” section of the AWS console. Choose “Users” and “Create new user”. Select a name for the user, and the “Programmatic access” option.
  2. On the following “Permissions” page, choose “Attach existing policies directly” and then press “Create policy”. Select the JSON tab and paste a policy like the example shown above, replacing my-example-opa-bucket with the name of your bucket.
  3. Once the policy has been created, it can be assigned to the user. With the user having been created, make sure to note down the AWS access key ID and the AWS secret access key, as they will be the credentials used for authentication.
Metadata Credentials
  1. Go to the “IAM” section of the AWS console. Choose “Roles” and “Create role”. For type, select “AWS service” and for use case, choose EC2, or wherever you’ll be running OPA.
  2. On the following “Permissions” page, choose “Create policy”. Select the JSON tab and paste a policy like the example shown above, replacing my-example-opa-bucket with the name of your bucket.
  3. Once the policy has been created, it can be assigned to the role.
  4. With the role created, go to the EC2 instance view. Select an instance where OPA will run and select “Actions” -> “Security” -> “Modify IAM role”. Select the role created in previous steps.
Testing Authentication

Use the [AWS CLI tools](command line tools) (see “Upload Bundle” below).

Upload Bundle

Bundle uploads to S3 are easily facilitated using the aws command in the AWS CLI tools.

  1. aws --profile=opa-service-account s3 cp bundle.tar.gz s3://my-example-opa-bucket/

Example OPA Configuration

Environment Credentials

With the environment variables AWS_REGION, AWS_ACCESS_KEY_ID and AWS_SECRET_ACCESS_KEY set, the following configuration will extract the credentials from the environment.

  1. services:
  2. s3:
  3. url: https://my-example-opa-bucket.s3.eu-north-1.amazonaws.com
  4. credentials:
  5. s3_signing:
  6. environment_credentials: {}
  7. bundles:
  8. authz:
  9. service: s3
  10. resource: bundle.tar.gz
Metadata Credentials

In order for this to work it is required that the permissions you created in the “Authentication” steps above are embedded in an IAM Role, which is then assigned to the EC2 instance hosting OPA.

  1. services:
  2. s3:
  3. url: https://my-example-opa-bucket.s3.eu-north-1.amazonaws.com
  4. credentials:
  5. s3_signing:
  6. metadata_credentials:
  7. aws_region: eu-north-1
  8. iam_role: my-opa-bucket-access-role
  9. bundles:
  10. authz:
  11. service: s3
  12. resource: bundle.tar.gz

Google Cloud Storage

OPA Bundle Support

FeatureSupported
Caching headersYes
Authentication methodsGCP Metadata Token
OAuth2 JWT Bearer Grant Type

Setup Instructions

  1. In the left pane menu, choose “Cloud Storage”. Click “New bucket”.
  2. Fill in the form according to your preferences (name, region, availability, etc).
  3. Once the bucket is created, you can press “Upload” to upload a test bundle. Clicking this will provide a link to the bundle which you can use in your OPA configuration.
  4. At this stage you can either choose to make the bucket public (by clicking “Permissions”) or to configure a service account for authenticated access.

Authentication

GCP Metadata Token Authentication

If your instance of OPA runs inside GCP, you’ll be able to authenticate using GCP metadata tokens. These tokens by default carry all the permissions granted to the default service account, so you might still want to create a dedicated service account for this purpose (see JWT Bearer Grant Type below).

JWT Bearer Grant Type

Use this for authenticating external clients, i.e. OPAs running outside the GCP environment.

  1. Search for “credentials” in the top search box and choose “Credentials - APIs and Services”.
  2. Click “Create Credentials” followed by “Service Account.”
  3. Fill in a name for the account and proceed to select roles.
  4. Choose “Storage Object Viewer” for read access and “Storage Object Creator” for write access (if scripted uploads is desired).
  5. Click the newly created service account and then the “Keys” tab. Press “Add Key” and either “Create new” or upload an existing one.
  6. If creating new, choose to download the private key in JSON format (not P12).
  7. Open the JSON file just downloaded and copy the PEM encoded value of the private_key attribute. This is the key you’ll use for your OPA configuration.
Testing Authentication

The easiest way of testing GCP metadata token or JWT bearer grant type authentication is simply to set up OPA with config for these and run the server.

Upload Bundle

Uploading a bundle is trivial with the gsutil command included with the Google Cloud SDK.

  1. gsutil cp bundle.tar.gz gs://<bucket-name>/

Example OPA Configuration

GCP Metadata Token Authentication
  1. services:
  2. gcs:
  3. url: https://storage.googleapis.com/storage/v1/b/${BUCKET_NAME}/o
  4. credentials:
  5. gcp_metadata:
  6. scopes:
  7. - https://www.googleapis.com/auth/devstorage.read_only
  8. bundles:
  9. authz:
  10. service: gcs
  11. # NOTE ?alt=media is required
  12. resource: 'bundle.tar.gz?alt=media'
Google Cloud Storage Bundle and JWT Bearer Authentication
  1. services:
  2. gcp:
  3. url: https://storage.googleapis.com/storage/v1/b/${BUCKET_NAME}/o
  4. credentials:
  5. oauth2:
  6. grant_type: jwt_bearer
  7. token_url: https://oauth2.googleapis.com/token
  8. signing_key: jwt_signing_key # references the key in `keys` below
  9. scopes:
  10. - https://www.googleapis.com/auth/devstorage.read_only
  11. additional_claims:
  12. aud: https://oauth2.googleapis.com/token
  13. iss: opa-client@my-account.iam.gserviceaccount.com
  14. bundles:
  15. authz:
  16. service: gcp
  17. # NOTE ?alt=media is required
  18. resource: 'bundle.tar.gz?alt=media'
  19. keys:
  20. jwt_signing_key:
  21. algorithm: RS256
  22. private_key: ${BUNDLE_SERVICE_SIGNING_KEY}

Azure Blob Storage

OPA Bundle Support

FeatureSupported
Caching headersYes
Authentication methodsOAuth2 Client Credentials,
OAuth2 Client Credentials JWT authentication

Note that for the time being, the Shared Key or Shared Access Signature (SAS) options are not supported.

Setup Instructions

  1. Any type of storage in Azure is grouped in Storage Accounts. If you have one already, skip to step 3.
  2. From the Azure console, select “Storage Accounts” followed by “New”. Fill in the form (name, region, etc) according to your preferences. One thing to note when selecting “account kind”, make sure to pick the Storage V2 (general purpose v2) option and not the legacy BlobStorage kind.
  3. With the storage account deployed, press “Go to resource” to create a new storage resource.
  4. Select “Containers” and press the plus sign to create a new storage container.
  5. Name your container and select access level. Choose “Private” to require authentication, or “Blob” to allow unauthenticated read access.
  6. Press “upload” and select the bundle from your local filesystem.
  7. Clicking the filename should bring up a properties window where the public URL to the bundle is included.

Authentication

  1. Go to Azure Active Directory.
  2. In the left menu, click “App Registrations” followed by “New Registration”. Name your app (client) amd leave the other options be. Click “Register”.
  3. Click “Certificates and Secrets”. Either create a secret to be used for OAuth2 Client Credentials or upload a certificate for OAuth2 Client Credentials JWT authentication.
  4. In the menu to the left, click “API permissions”. Click “Add a permission”. Choose “Azure Storage” and check the “user_impersonation” checkbox.
  5. Click “Add admin consent for Default Directory”. Answer Yes on the followup question.
  6. Navigate back to your storage account. Click “Access Control (IAM)”. Click “Add role assignments”.
  7. Select the “Storage Blob Data Contributor” role. Leave “Assign access to” as “User, group or service principal”. Search and select the name of the app created in step 2.
  8. Configuration is now complete. Go back to “App Registrations” in the Active Directory view to check details like tenant ID, application ID and endpoints. You’ll need those when configuring OPA (see Example Configuration below).
Testing Authentication

Use Curl to test client authentication with a secret.

  1. curl --silent \
  2. --data "grant_type=client_credentials&client_id=$CLIENT_ID&client_secret=$CLIENT_SECRET&scope=https://storage.azure.com/.default" \
  3. "https://login.microsoftonline.com/$TENANT_ID/oauth2/v2.0/token"

Upload Bundle

Uploading bundles to Azure Blob storage is easily done using the azcopy tool. Make sure to first properly authorize the user to be able to upload to Blob storage.

By now you should be able to login interactively using azcopy login --tenant-id <Active Directory tenant ID>. Since you’ll most likely will want to log in from scripts (to upload bundles programmatically), you should however create an Azure AD application, and a service principal to do so. Good news! If you’ve followed the Authentication steps above, you already have one.

Uploading bundle using client secret authentication

  1. AZCOPY_SPA_CLIENT_SECRET='<application_client_secret>' azcopy login \
  2. --service-principal \
  3. --tenant-id <tenant-id> \
  4. --application-id <application-id>
  5. azcopy copy bundle.tar.gz https://<storage-account-id>.blob.core.windows.net/<container-id>/bundle.tar.gz

Uploading bundle using client certificate authentication

  1. AZCOPY_SPA_CERT_PASSWORD='<client_cert_password>' azcopy login \
  2. --service-principal \
  3. --tenant-id <tenant-id> \
  4. --certificate-path <path-to-certificate-file> --tenant-id <tenant-id>
  5. azcopy copy bundle.tar.gz https://<storage-account-id>.blob.core.windows.net/<container-id>/bundle.tar.gz

Uploading bundle using Curl

  1. token=$(curl --silent \
  2. --data "grant_type=client_credentials&client_id=$CLIENT_ID&client_secret=$CLIENT_SECRET&scope=https://storage.azure.com/.default" \
  3. "https://login.microsoftonline.com/$TENANT_ID/oauth2/v2.0/token" | jq -r .access_token)
  4. curl --silent \
  5. -X PUT \
  6. --data-binary "@bundle.tar.gz" -H "X-Ms-Version: 2020-04-08" -H "Authorization: Bearer $token" \
  7. https://styra.blob.core.windows.net/opa/bundle.tar.gz

Example OPA Configuration

Azure Blob Storage Bundle and Client Credentials Authentication
  1. services:
  2. blob:
  3. url: https://my-storage-account.blob.core.windows.net
  4. headers:
  5. # This header _must_ be present in all authenticated requests
  6. x-ms-version: "2020-04-08"
  7. credentials:
  8. oauth2:
  9. token_url: "https://login.microsoftonline.com/${TENANT_ID}/oauth2/v2.0/token"
  10. client_id: "${CLIENT_ID}"
  11. client_secret: "${CLIENT_SECRET}"
  12. scopes:
  13. - https://storage.azure.com/.default
  14. bundles:
  15. authz:
  16. service: blob
  17. resource: my-container/bundle.tar.gz

Note that the $CLIENT_ID is what is referred to as the “Application ID” inside your Azure account.

Azure Blob Storage Bundle and Client Credentials JWT Authentication
  1. keys:
  2. blob_key:
  3. algorithm: RS256
  4. private_key: "${PRIVATE_KEY_PEM}"
  5. services:
  6. blob:
  7. url: https://my-storage-account.blob.core.windows.net
  8. headers:
  9. # This header _must_ be present in all authenticated requests
  10. x-ms-version: "2020-04-08"
  11. credentials:
  12. oauth2:
  13. token_url: "https://login.microsoftonline.com/${TENANT_ID}/oauth2/v2.0/token"
  14. signing_key: blob_key
  15. thumbprint: "8F1BDDDE9982299E62749C20EDDBAAC57F619D04"
  16. include_jti_claim: true
  17. scopes:
  18. - https://storage.azure.com/.default
  19. additional_claims:
  20. aud: "https://login.microsoftonline.com/${TENANT_ID}/oauth2/v2.0/token"
  21. iss: "${CLIENT_ID}"
  22. sub: "${CLIENT_ID}"
  23. bundles:
  24. authz:
  25. service: blob
  26. resource: opa/bundle.tar.gz

Note that the $CLIENT_ID is what is referred to as the “Application ID” inside your Azure account. Also note in particular how the thumbprint property is required for Azure. The value expected here can be found under “Certificates and Secrets” in your application’s configuration.

Bundles - 图1

Certificate thumbprint

Nginx

Nginx offers a simple but competent bundle server for those who prefer to host their own. A great choice or for local testing.

FeatureSupported
Caching headersYes
Authentication methodsBearer Token 1
OAuth2 Client Credentials JWT authentication 2

1Nginx does not support bearer token authentication, but it does support basic auth. This can be achieved by setting services[_].credentials.bearer.scheme to Basic in the OPA configuration, and simply provide the base64 encoded credentials as the token.
2Only available with Nginx Plus.

Upload Bundle

Either use the nginx-upload-module or upload bundles out-of-band with SSH or similar.

Example OPA Configuration

  1. services:
  2. nginx:
  3. url: https://my-nginx.example.com
  4. credentials:
  5. bearer:
  6. token: dGVzdGluZzp0ZXN0aW5n
  7. scheme: Basic
  8. bundles:
  9. authz:
  10. service: nginx
  11. resource: /bundle.tar.gz