Open dave-tomkinson opened 1 week ago
@dave-tomkinson Thanks for the bug report. Would you be able to provide a model with column names and configs for one of the models that fails? You can redact the logic that goes into each column if you prefer. I need to know things like whether the columns are quoted or not, whether they are lowercase or uppercase, etc. I suspect that the SDK returns columns that do not match the case, or which are/aren't quoted as expected. This could cause dbt to attempt to add a column that's already there because of a scenario like "id" <> "Id"
. I'll setup a skeleton of a reproduction in the meantime, but nothing beats real data. Thanks in advance!
One additional note, I'm making the assumption that this is an incremental model with on_schema_change
set to either append_new_columns
or sync_all_columns
, hence the "column already exists" error. Please let me know if that's incorrect.
Is this a new bug in dbt-redshift?
Current Behavior
When I switch dbt cloud to versionless and I build with no flags set then the build works, but with Warnings to set the flag
restrict_direct_pg_catalog_access
When I set the flag
restrict_direct_pg_catalog_access
to True the build fails, with a number of tables (probably around 10%) FailingMostly with one of these Errors:
Setting it to False or removing the flag means the build works again.
Expected Behavior
The build to succeed with no Warnings.
Steps To Reproduce
In dbt cloud with the project version set to
versionless
Set the following flag indbt_project.yml
Run
dbt build
Relevant log output
No response
Environment
Additional Context
In the above