添加Bigtable外部表至BigQuery时遇“Not found”错误求助

Let's walk through the most likely fixes for your issue, starting with the easiest to verify:

1. Correct Your Bigtable URI Format #

The first red flag is your URI structure—you’re missing the critical bigtable. subdomain in the base URL. Your current URI:

https://googleapis.com/bigtable/projects/[PROJECT_ID]/instances/[INSTANCE_ID]/tables/[TABLE_NAME]

This is invalid because Google’s services can’t route the request to the Bigtable API without the subdomain. Update it to either of these valid formats:

• Full HTTPS URI: https://bigtable.googleapis.com/bigtable/projects/[PROJECT_ID]/instances/[INSTANCE_ID]/tables/[TABLE_NAME]
• Simplified protocol format (easier to avoid typos): bigtable://projects/[PROJECT_ID]/instances/[INSTANCE_ID]/tables/[TABLE_NAME]

Fixing this alone will resolve most "Not found" errors for this scenario.

2. Verify IAM Permissions for BigQuery’s Service Account #

Even with a correct URI, BigQuery needs explicit permissions to access your Bigtable instance. Here’s what you need to set up:

• Grant Bigtable data access: Find your BigQuery service account (format: service-PROJECT_NUMBER@gcp-sa-bigquery.iam.gserviceaccount.com) and assign the roles/bigtable.dataReader role to it on your Bigtable instance or table. This lets BigQuery read data from Bigtable.
• Ensure table creation permissions: The account you’re using to create the BigQuery external table needs at minimum bigquery.tables.create and bigtable.tables.get permissions (or broader roles like roles/bigquery.admin or roles/bigtable.admin).

To locate your BigQuery service account:

1. Navigate to your BigQuery project’s IAM page.
2. Look for the account starting with service- followed by your project number, ending with @gcp-sa-bigquery.iam.gserviceaccount.com.
3. Copy this email, go to your Bigtable instance’s IAM page, and add the Bigtable Data Reader role for that account.

3. Development Instance Impact (Or Lack Thereof) #

Development instances don’t restrict integration with BigQuery—they function the same as production instances for this use case. The instance type isn’t causing your error, but it’s worth confirming your BigQuery dataset and Bigtable instance are in the same region (cross-region access works, but same-region avoids latency and potential connectivity quirks).

Final Test Steps #

After fixing the URI and permissions:

• Try creating the external table again via the BigQuery UI, making sure you select "Cloud Bigtable" as the external data source and paste the corrected URI.
• If using the bq CLI, run this command (replace placeholders):

  bq mk --external_table_definition=bigtable://projects/[PROJECT_ID]/instances/[INSTANCE_ID]/tables/[TABLE_NAME] [DATASET].[TABLE]
  

内容的提问来源于stack exchange，提问作者Mike Sukmanowsky