I try to use sling to transfer tables from sqlserver to fabric using bulk loading. If I login using my interactive login (az login) everything works, but i need to use a SPN to connecti to the fabric WH. Using an spn I am not able to get the COPY INTO command to work properly. The error message is: mssql: Content of directory on path 'https://XXXXXX.dfs.core.windows.net/dls-edp-test-01/slingtest/sling_temp/sling.dimsupplier_sling_tmp/1768995259092/*' cannot be listed. More information: I have configured the staging like this: abfs_endpoint: XXXXXXX.dfs.core.windows.net abfs_filesystem: dls-edp-test-01 abfs_parent: slingtest The upload of the parquet-file to the staging area works. I can restore the deleted files with azure storage explorer, and everything seems fine. However, the COPY INTO command gives the error above, even if I run it manually from SSMS logged in with my SPN. If I change the COPY INTO command by changing the URL from XXXXXXX.dfs.core.windows.net to XXXXXXX.dfs.blob.windows.net, the COPY command succeeds. If I try using the blob version as the abfs_endpoint the upload of the parquet file fails. So, it seems I need to use dfs when writing to the staging area, but blob when reading. Any ideas?

hey @stegus64 looked into this, and added a new property

copy_into_endpoint

that you can specify in your env.yaml entry. Should be availabe next release (see docs https://docs.slingdata.io/connections/database-connections/fabric#copy-into-endpoint-issues). Planning to release 1.5.5 tomorrow, let me know how that works out. Thanks for raising.

Hello! This sounds excellent! I have actually created a fork of sling and made exactly the same change. So I know this will work. Thank you for your quick attention!

hey @stegus64 actually, after thinking about it, this logic was removed. The user will need to explicitly specify the

copy_into_endpoint

endpoint to replace, so that there are no surprises.

ok, that sounds reasonable. Maybe you should update the documentation to make it less confusing?

yes, already updated

By the way, I tried the new 1.5.5 release, and it works perfectly after configuring copy_into_endpoint. Good job!