This is the stable branch for Tag Engine. Tag Engine v2 is a flavor of Tag Engine that is compatible with Data Catalog and is hosted on Cloud Run. It supports user authentication and role based access control. Customers who have multiple teams using BigQuery and Cloud Storage can authorize each team to tag only their data assets using separate Tag Creator service accounts (more on that later).
If you are looking for Dataplex support, please check out the dataplex branch in this repo.
Tag Engine v2 is an open-source extension to Data Catalog on Google Cloud, which is now part of the Dataplex product suite. Tag Engine automates the tagging of BigQuery tables and views as well as data lake files in Cloud Storage. You create tag configurations that specify how to populate the various fields of a tag template through SQL expressions or static values. Tag Engine runs the configurations either on demand or on a schedule to create, update or delete the tags.
This README file contains deployment steps, testing procedures, and code samples. It is organized into five sections:
- Part 1: Deploying Tag Engine v2
- Part 2: Testing your Tag Engine API Setup
- Part 3: Testing your Tag Engine UI Setup
- Part 4: Troubleshooting
- Part 5: Next Steps
Tag Engine v2 comes with two Cloud Run services. One service is for the API (tag-engine-api
) and the other is for the UI (tag-engine-ui
).
Both services use access tokens for authorization. The API service expects the client to pass in an access token when calling the API functions (gcloud auth print-identity-token
) whereas the UI service uses OAuth to authorize the client from the front-end. Note that a client secret file is required for the OAuth flow.
Follow the steps below to deploy Tag Engine with Terraform.
Alternatively, you may choose to deploy Tag Engine with gcloud commands instead of running the Terraform.
-
Create (or designate) two service accounts:
- A service account that runs the Tag Engine Cloud Run services (both API and UI). This account is referred to as
TAG_ENGINE_SA
. - A service account that sources the metadata from BigQuery or Cloud Storage, and then performs the tagging in Data Catalog. This account is referred to as
TAG_CREATOR_SA
.
See Creating Service Accounts for more details.
Why do we need two different service accounts? The key benefit of decoupling them is to allow individual teams to have their own Tag Creator SA. This account has permissions to read specific data assets in BigQuery and Cloud Storage. For example, the Finance team can have a different Tag Creator SA from the Finance team if they own different data assets. The Tag Engine admin then links each invoker account (either service or user) to a specific Tag Creator SA. Invoker accounts call Tag Engine through either the API or UI. This allows the Tag Engine admin to run and maintain a single instance of Tag Engine, as opposed to one instance per team.
- A service account that runs the Tag Engine Cloud Run services (both API and UI). This account is referred to as
-
Create an OAuth client:
Open API Credentials.
Click on Create Credentials and select OAuth client ID and choose the following settings:
Application type: web application
Name: tag-engine-oauth
Authorized redirects URIs: Leave this field blank for now.
Click Create
Download the credentials aste_client_secret.json
and place the file in the root of thedatacatalog-tag-engine
directoryNote: The client secret file is required for establishing the authorization flow from the UI.
-
Make a copy of
datacatalog-tag-engine/tagengine.ini.tpl
naming the new copydatacatalog-tag-engine/tagengine.ini
. -
Open
datacatalog-tag-engine/tagengine.ini
and set the following variables in this file:TAG_ENGINE_SA TAG_CREATOR_SA TAG_ENGINE_PROJECT TAG_ENGINE_REGION FIRESTORE_PROJECT FIRESTORE_REGION FIRESTORE_DATABASE BIGQUERY_REGION FILESET_REGION SPANNER_REGION ENABLE_AUTH OAUTH_CLIENT_CREDENTIALS ENABLE_TAG_HISTORY TAG_HISTORY_PROJECT TAG_HISTORY_DATASET ENABLE_JOB_METADATA JOB_METADATA_PROJECT JOB_METADATA_DATASET
A couple of notes:
-
The variable
ENABLE_AUTH
is a boolean. When set toTrue
, Tag Engine verifies that the end user is authorized to useTAG_CREATOR_SA
prior to processing their tag requests. This is the recommended value. -
The
tagengine.ini
file also has two additional variables,INJECTOR_QUEUE
andWORK_QUEUE
. These determine the names of the cloud task queues. You do not need to change them. If you change their name, you need to also change them in thedeploy/variables.tf
.
-
-
Create a new GCS bucket for CSV imports. Remember GCS bucket names are globally unique. For example:
gsutil mb gs://$(gcloud config get-value project)-csv-import
-
Set the Terraform variables:
Open
deploy/variables.tf
and change the default value of each variable.
Save the file.
Alternatively, create a new file, nameddeploy/terrform.tfvars
and specify your variables values there. -
Run the Terraform scripts:
NOTE: The terraform script will run with the default credentials currently configured on your system. Make sure that your current user has the required permissions to make changes to your project(s), or set new credentials using the
GOOGLE APPLICATION_CREDENTIALS
environment variable.cd deploy terraform init terraform plan terraform apply
When the Terraform finishes running, it should output two URIs. One for the API service (which looks like this https://tag-engine-api-xxxxxxxxxxxxx.a.run.app) and another for the UI service (which looks like this https://tag-engine-ui-xxxxxxxxxxxxx.a.run.app).
-
Create the sample
data_governance
tag template:git clone https://github.com/GoogleCloudPlatform/datacatalog-templates.git cd datacatalog-templates python create_template.py $DATA_CATALOG_PROJECT $DATA_CATALOG_REGION data_governance.yaml
The previous command creates the
data_governance
tag template in the$DATA_CATALOG_PROJECT
and$DATA_CATALOG_REGION
. -
Grant permissions to invoker account (user or service)
Depending on how you are involving the Tag Engine API, you'll need to grant permissions to either your service account or user account (or both).
If you'll be invoking the Tag Engine API with a user account, authorize your user account as follows:
gcloud auth login export INVOKER_USER_ACCOUNT="username@example.com" gcloud iam service-accounts add-iam-policy-binding $TAG_CREATOR_SA \ --member=user:$INVOKER_USER_ACCOUNT --role=roles/iam.serviceAccountUser --project=$DATA_CATALOG_PROJECT gcloud run services add-iam-policy-binding tag-engine-api \ --member=user:$INVOKER_USER_ACCOUNT --role=roles/run.invoker \ --project=$TAG_ENGINE_PROJECT --region=$TAG_ENGINE_REGION
If you are invoking the Tag Engine API with a service account, authorize your service account as follows:
export INVOKER_SERVICE_ACCOUNT="tag-engine-invoker@<PROJECT>.iam.gserviceaccount.com" gcloud iam service-accounts add-iam-policy-binding $TAG_CREATOR_SA \ --member=serviceAccount:$INVOKER_SERVICE_ACCOUNT --role=roles/iam.serviceAccountUser gcloud run services add-iam-policy-binding tag-engine-api \ --member=serviceAccount:$INVOKER_SERVICE_ACCOUNT --role=roles/run.invoker \ --region=$TAG_ENGINE_REGION
Very important: Tag Engine requires that these roles be directly attached to your invoker account(s).
-
Generate an IAM token (aka Bearer token) for authenticating to Tag Engine:
If you are invoking Tag Engine with a user account, run
gcloud auth login
and authenticate with your user account. If you are invoking Tag Engine with a service account, setGOOGLE_APPLICATION_CREDENTIALS
.export IAM_TOKEN=$(gcloud auth print-identity-token)
-
Create your first Tag Engine configuration:
Tag Engine uses configurations (configs for short) to define tag requests. There are several types of configs from ones that create dynamic table-level tags to ones that create tags from CSV. You'll find several example configs in the
examples/configs/
subfolders.For now, open
examples/configs/dynamic_table/dynamic_table_ondemand.json
and update the project and dataset values in this file to match your Tag Engine and BigQuery environments.cd datacatalog-tag-engine export TAG_ENGINE_URL=$SERVICE_URL curl -X POST $TAG_ENGINE_URL/create_dynamic_table_config -d @examples/configs/dynamic_table/dynamic_table_ondemand.json \ -H "Authorization: Bearer $IAM_TOKEN"
Note:
$SERVICE_URL
should be equal to your Cloud Run URL fortag-engine-api
.The output from the previous command should look similar to:
{"config_type":"DYNAMIC_TAG_TABLE","config_uuid":"facb59187f1711eebe2b4f918967d564"}
-
Run your first job:
Now that we have created a config, we need to trigger it in order to create the tags. A Tag Engine job is an execution of a config. In this step, you execute the dynamic table config using the config_uuid from the previous step.
Note: Before running the next command, please update the
config_uuid
with your own value.curl -i -X POST $TAG_ENGINE_URL/trigger_job \ -d '{"config_type":"DYNAMIC_TAG_TABLE","config_uuid":"facb59187f1711eebe2b4f918967d564"}' \ -H "Authorization: Bearer $IAM_TOKEN"
The output from the previous command should look similar to:
{ "job_uuid": "069a312e7f1811ee87244f918967d564" }
If you enabled job metadata in
tagengine.ini
, you can optionally pass a job metadata object to the trigger_job call. This gets stored in BigQuery, along with the job execution details. Please note that the job metadata option is not required, you can skip this step:curl -i -X POST $TAG_ENGINE_URL/trigger_job \ -d '{"config_type":"DYNAMIC_TAG_TABLE","config_uuid":"c255f764d56711edb96eb170f969c0af","job_metadata": {"source": "Collibra", "workflow": "process_sensitive_data"}}' \ -H "Authorization: Bearer $IAM_TOKEN"
The job metadata parameter gets written into a BigQuery table that is associated with the job_uuid.
-
View your job status:
Note: Before running the next command, please update the
job_uuid
with your value.curl -X POST $TAG_ENGINE_URL/get_job_status -d '{"job_uuid":"069a312e7f1811ee87244f918967d564"}' \ -H "Authorization: Bearer $IAM_TOKEN"
The output from this command should look like this:
{ "job_status": "SUCCESS", "task_count": 1, "tasks_completed": 1, "tasks_failed": 0, "tasks_ran": 1 }
Open the Data Catalog UI and verify that your tag was successfully created. If your tag is not there or if you encounter an error with the previous commands, open the Cloud Run logs for the
tag-engine-api
service and investigate.
-
Set the authorized redirect URI and add authorized users:
-
Re-open API Credentials
-
Under OAuth 2.0 Client IDs, edit the
tag-engine-oauth
entry which you created earlier. -
Under Authorized redirect URIs, add the URI: https://tag-engine-ui-xxxxxxxxxxxxx.a.run.app/oauth2callback
-
Replace xxxxxxxxxxxxx in the URI with the actual value from the Terraform. This URI will be referred to below as the
UI_SERVICE_URI
. -
Open the OAuth consent screen page and under the Test users section, click on add users.
-
Add the email address of each user for which you would like to grant access to the Tag Engine UI.
-
-
Grant permissions to your invoker user account(s):
export INVOKER_USER_ACCOUNT="username@example.com"` gcloud iam service-accounts add-iam-policy-binding $TAG_CREATOR_SA \ --member=user:$INVOKER_USER_ACCOUNT --role=roles/iam.serviceAccountUser
-
Open a browser window
-
Navigate to
UI_SERVICE_URI
-
You should be prompted to sign in with Google
-
Once signed in, you will be redirected to the Tag Engine home page (i.e.
UI_SERVICE_URI
/home) -
Enter your template id, template project, and template region
-
Enter your
TAG_CREATOR_SA
as the service account -
Click on
Search Tag Templates
to continue to the next step -
Create a tag configuration by selecting one of the options from this page.
If you encounter a 500 error, open the Cloud Run logs for
tag-engine-ui
to troubleshoot.
There is a known issue with the Terraform. If you encounter the error The requested URL was not found on this server
when you try to create a configuration from the API, the issue is that the container didn't build correctly. Try to rebuild and redeploy the Cloud Run API service with this command:
cd datacatalog-tag-engine
gcloud run deploy tag-engine-api \
--source . \
--platform managed \
--region $TAG_ENGINE_REGION \
--no-allow-unauthenticated \
--ingress=all \
--memory=4G \
--timeout=60m \
--service-account=$TAG_ENGINE_SA
Then, call the ping
endpoint as follows:
curl $TAG_ENGINE_URL/ping -H "Authorization: Bearer $IAM_TOKEN"
You should see the following response:
Tag Engine is alive
-
Explore additional API methods and run them through curl commands:
Open
examples/unit_test.sh
and go through the different methods for interracting with Tag Engine, includingconfigure_tag_history
,create_static_asset_config
,create_dynamic_column_config
, etc. -
Explore the script samples:
There are multiple test scripts in Python in the
examples/scripts
folder. These are intended to help you get started with the Tag Engine API.Before running the scripts, open each file and update the
TAG_ENGINE_URL
variable on line 11 with your own Cloud Run service URL. You'll also need to update the project and dataset values which may be in the script itself or in the referenced json config file.Here are some of the scripts you can look at and run:
python configure_tag_history.py python create_static_config_trigger_job.py python create_dynamic_table_config_trigger_job.py python create_dynamic_column_config_trigger_job python create_dynamic_dataset_config_trigger_job.py python create_import_config_trigger_job.py python create_export_config_trigger_job.py python list_configs.py python read_config.py python purge_inactive_configs.py
-
Explore sample workflows:
The
extensions/orchestration/
folder contains some sample workflows implemented in Cloud Workflow. Thetrigger_tag_export.yaml
andtrigger_tag_export_import.yaml
show how to orchestrate Tag Engine jobs. To run the workflows, enable the Cloud Workflows API (workflows.googleapis.com
) and then follow these steps:gcloud workflows deploy orchestrate-jobs --location=$TAG_ENGINE_REGION \ --source=trigger_export_import.yaml --service-account=$CLOUD_RUN_SA gcloud workflows run trigger_export_import --location=$TAG_ENGINE_REGION
In addition to the Cloud Workflow examples, there are two examples for Airflow in the same folder,
dynamic_tag_update.py
andpii_classification_dag.py
.
- Create your own Tag Engine configs with the API and/or UI.
- Open new issues if you encounter bugs or would like to request a new feature or extension.