Replies: 4 comments 3 replies
-
|
Hi, can you check if the secret is injected into the duckdb session by BTW, if you have access to server log, maybe enable |
Beta Was this translation helpful? Give feedback.
-
|
I suspect it's something to do with the join between a PG table and a duckdb-backed view. We have 4 other extracts which extract duckdb-backed views only and they are working fine, the only difference between those and this one is that this one joins to a PG table... |
Beta Was this translation helpful? Give feedback.
-
|
Have upgraded the version of pg_duckdb to v1.1.1 that we're running (still on pg16), and get a little bit more detail from the error. It's saying no credentials provided in the error message in Tableaus jobs UI along with the standard HTTP GET info. So the user has credentials through USER MAPPING, we have a view which is a Tableau is trying to query a PG table, joining the duckdb-backed view to the PG table, and somewhere, the USER MAPPING credentials aren't propagating to allow duckdb to query the parquet. |
Beta Was this translation helpful? Give feedback.
-
|
This is very much a tableau thing. We've swapped to using an embedded JDBC postgres connection and the join works and executes fine. So there's something that tableau does with the native connector it ships with that fudges up duckdb. |
Beta Was this translation helpful? Give feedback.
Uh oh!
There was an error while loading. Please reload this page.
-
We're having a fairly odd issues with pg_duckdb on pg16.
We've got a table in our data lake (S3) which we expose via a view in our pg16 warehouse as a view (duckdb query) over the parquet file.
For a given user, with a USER MAPPING with key/secret, we can query that view as the user perfectly fine from tools like DataGrip, or psycopg in python, or PGAdmin. As well, we can query the table from Tableau Desktop perfectly fine using the normal Tableau postgres connector, or the postgres JDBC connector.
When we move that workbook to our tableau server though, everything seems to change. Same user, Same warehouse behind the scenes, with the same USER MAPPING and keys. But when the extract tries to run the query to pull the views data, we get a 403 forbidden HTTP GET error with what appears to be an incorrect endpoint structure for the S3 bucket:
https://our-bucket-name.s3.amazonaws.com/path/file.parquetis the path the extract process complains about, but pretty sure that it should behttps://our-bucket-name.s3.region.amazonaws.com/path/file.parquet. We've tried setting the endpoint on the USER MAPPING and we already set the region on it. The tableau server is in the same VPC and subnet as the postgres warehouse so I don't think it's down to network path at all.Any ideas how Tableau might be interacting with it's postgres connection in a way that alters how the duckdb query under the hood would operate? Some sort of changes to the session that could cause something like this?
Beta Was this translation helpful? Give feedback.
All reactions