What causes DELTA_SCHEMA_CHANGED?

A column was added or removed from the table between job start and commit. Another job ran with `mergeSchema=true` and added columns, changing the committed schema. A dbt model ran CREATE OR REPLACE TABLE with a modified column list during the write window. Upstream Autoloader schema inference detected new fields and updated the table schema mid-stream

How do I fix DELTA_SCHEMA_CHANGED?

Retry the failed job — on retry it will read the current schema and proceed correctly.. Enable `spark.databricks.delta.schema.autoMerge.enabled = true` on the cluster if controlled schema evolution is acceptable.. Coordinate schema changes across jobs: use a maintenance window or job dependency to prevent overlap.. For streaming jobs, set `mergeSchema` explicitly in the writeStream options rather than relying on cluster-level defaults.

High severitysql

Power BI Refresh Error:
DELTA_SCHEMA_CHANGED

What does this error mean?

The schema of the Delta table changed between when the write transaction started and when it attempted to commit. The transaction is aborted because its data does not match the new schema.

Common causes

1A column was added or removed from the table between job start and commit
2Another job ran with `mergeSchema=true` and added columns, changing the committed schema
3A dbt model ran CREATE OR REPLACE TABLE with a modified column list during the write window
4Upstream Autoloader schema inference detected new fields and updated the table schema mid-stream

How to fix it

1Retry the failed job — on retry it will read the current schema and proceed correctly.
2Enable `spark.databricks.delta.schema.autoMerge.enabled = true` on the cluster if controlled schema evolution is acceptable.
3Coordinate schema changes across jobs: use a maintenance window or job dependency to prevent overlap.
4For streaming jobs, set `mergeSchema` explicitly in the writeStream options rather than relying on cluster-level defaults.

Frequently asked questions

Can I prevent this error without serializing all writes?

Partial mitigation: enable autoMerge so new columns are accepted automatically. This prevents conflicts on additive schema changes but not on column removal or type changes.