Unable to update DLT-based materialized view if clustering key is missing

[Malthe]

If we set up a materialized view with a clustering key, and then update the definition such that this key is no longer part of the table, Databricks complains:

Run ALTER TABLE ... CLUSTER BY ... to repair Delta clustering metadata.

But this is not possible and you can't do a full refresh either. One has to DROP the materialized view and recreate it (i.e. manual repair).

It should be possible to remedy this situation by doing a full refresh.

[K_Anudeep · Accepted solution]

Hello @Malthe ,

Currently, there is no supported way to repair broken clustering metadata in Delta materialised views if you remove the clustering key from the definition, other than dropping and recreating the materialised view. Additionally, a full refresh cannot resolve the underlying Delta clustering metadata inconsistency when the clustering key is removed or changed, as fullRefresh is meant for recomputing the data, not rewriting the MV’s underlying Delta table properties (such as clustering keys).

I have checked this internally, and it appears to be a known issue that will be resolved and rolled out by the end of January 2026.