1. Help Center
  2. Analytics SDKs
  3. Unified Digital Model

Snowplow Unified DBT Incremental Manifest Reset on Excluded Model Runs

Unified Digital Model
Edwin Mejias  
Edited

Summary

Customers using Snowplow’s Unified DBT package may encounter an issue where the snowplow_unified_incremental_manifest table is unexpectedly reset during runs that should have excluded incremental models.

This article documents a known compatibility issue between DBT Core and the Snowplow Unified model setup that can result in manifest corruption, leading to partial or missing data in downstream derived tables.

Affected Products

  • DBT Core ≥ 1.7.0
  • Snowplow Unified DBT Model
  • Snowplow Utils v0.17.2

Symptoms

  • models_matched_from_manifest evaluates to zero despite valid models being selected.
  • snowplow_unified_incremental_manifest table is unexpectedly recreated.
  • Incremental logic is skipped, resulting in missing or overwritten data.
  • This occurs even when using the --exclude tag:snowplow_unified_incremental flag.
  • snowplow__allow_refresh: false setting is not respected in certain conditions.

Root Cause

This issue stems from a DBT Core bug (dbt-labs/dbt-core#9789), where partial parsing creates an outdated or inaccurate model manifest. As a result, dbt does not recognize that some models were previously run or tagged correctly, leading to an incorrect value for models_matched_from_manifest.

This caused macros such as get_run_limits.sql in the snowplow-utils package to interpret the manifest as empty, thereby triggering a rebuild of the snowplow_unified_incremental_manifest table — even when it shouldn't.

Workaround / Resolution

To prevent the issue, disable partial parsing during your dbt runs:

dbt run --no-partial-parse

This ensures dbt fully rebuilds its internal model manifest and correctly detects model tags and dependencies.

Additional Best Practices

  • Always pin both your snowplow_utils and dbt versions explicitly in your packages.yml and deployment environment.
  • Monitor the size of the snowplow_unified_incremental_manifest table to detect resets or empty states.
  • Add logging to capture values such as models_in_run and models_matched_from_manifest during test runs.
  • Consider disabling --partial-parse in production pipelines when making structural changes or upgrades to dbt packages.

Support Scope

Note that Snowplow does not maintain the dbt-core framework and cannot guarantee fixes for upstream bugs. All assistance provided in this area is on a best-effort basis.

For persistent dbt issues outside of Snowplow models, we recommend contacting dbt Support directly.