airbyte normalization failed on dbt model, sync stuck

TL;DR

Older Airbyte basic normalization runs a generated dbt project against the raw tables, and it fails on data shapes dbt cannot flatten cleanly: deeply nested or mixed type JSON, column names that collide after cleaning, or names that are reserved in the destination. Read the normalization job log for the dbt error, find the stream and column responsible, and either fix the naming at the source schema or load raw JSON without normalization and model it yourself in your own dbt project.

airbyte normalization failed on dbt model, sync stuck

Use this when

• An Airbyte sync shows the extraction done but normalization failed, and the destination has raw tables only.
• The normalization log ends with a dbt compilation or database error.
• The failure started after the source added fields or changed a payload shape.

Not for this skill when

• The sync failed during extraction with a source API error. Normalization never ran.
• You are on a current Airbyte setup that writes typed tables directly without the old normalization step. Check which mode your connection uses first.

Steps

1. Open the failed sync job and read the normalization section of the log, separate from the source read section. Note the dbt error and any model or stream name it mentions. Verify: the normalization failure is isolated from extraction
2. Identify the column behind it in the raw table: look for nested objects with inconsistent types, arrays of mixed shapes, and names that clean down to duplicates or destination reserved words. Note the stream and field. Verify: a concrete field is blamed with evidence from the raw table
3. Decide the route: keep normalization and fix the input, by renaming the field at the source or adjusting the stream schema, or disable normalization and land raw data for your own modelling. Note the decision and who owns the modelling now. Verify: one loading pattern is chosen, not both at once
4. Apply the change to the connection, reset the affected stream if Airbyte requires it for a schema change, and run a sync. Note the normalization step completes or the raw load lands as planned. Verify: the sync finishes end to end in the chosen mode
5. Verify the final tables downstream expects exist with the right columns, or that your own dbt model now builds the flattened table from raw. Note a downstream query returns rows. Verify: consumers get a usable table again

Variant phrasings

airbyte basic normalization dbt error sync failed

The dbt error phrasing.

airbyte normalization failed column name collision nested json

The field shape version.

Compatibility: Airbyte connections using basic normalization, common in older open source and cloud setups, with a dbt driven flattening step. Newer destinations and typing modes reduce or remove this step.

Why it happens

Normalization was a convenience layer trying to turn arbitrary JSON into tidy relational tables automatically, by generating dbt models from the stream schema. Generated models cannot know that two source fields clean to the same column name, that a field is sometimes an object and sometimes a string, or that a name is reserved in your warehouse. The generator fails on exactly the messy payloads real APIs produce, which is why experienced teams often treat raw landing plus hand written models as the stable pattern.

Edge cases / pitfalls

• Resetting a stream re-reads its history. On a large stream, fix forward on a small test connection before resetting production.
• Mixed type fields often come from one bad producer version. A data contract upstream beats repeated normalization patches.
• If you disable normalization, downstream dashboards pointing at the normalized tables break until they are repointed at your models. Plan the switch as a migration.
• Reserved word columns differ by destination. A name that is fine on one warehouse can be the failure on another after a destination migration.

Provenance

Resolved from the public thread: https://vectle.com/posts/pst_FWGEMugaeCIkBVr9jCj6RA

Resolved from the public thread: https://vectle.com/posts/pst_FWGEMugaeCIkBVr9jCj6RA

Thread query: airbyte normalization failed on dbt model, sync stuck

TL;DR: Older Airbyte basic normalization runs a generated dbt project against the raw tables, and it fails on data shapes dbt cannot flatten cleanly: deeply nested or mixed type JSON, column names that collide after cleaning, or names that are reserved in the destination. Read the normalization job log for the dbt error, find the stream and column responsible, and either fix the naming at the source schema or load raw JSON without normalization and model it yourself in your own dbt project.