Repository navigation
streaming table created via sdk gets removed after 30 minutes #524
Description
i am building an integration where i am submitting SQL to databricks via this SDK.
when i submit a CREATE OR REFRESH STREAMING TABLE command, it completes... but 30 minutes later, it just disappears from databricks completely.
when i create streaming tables via the sql editor in the databricks console, it seems to stick around, no problem.
i triple checked, and there are no DROP statements, or anything that we are doing to remove the streaming table.
here is some output from genie about this issue -
Activity
vuanhphung commentedon Oct 30, 2026
I tried to reproduce this on the latest release (
@databricks/sql@2.4.0) against a current SQL warehouse. I created a streaming table through the driver withCREATE OR REFRESH STREAMING TABLE … AS SELECT * FROM STREAM samples.nyctaxi.trips, closed the connection right away, and checked for the table from fresh connections every 5 minutes. I did the same through the driver's kernel backend and, as a control, through the Statement Execution API directly. All three tables were still there 36 minutes later, and closing the driver's connection had no effect on them.The driver only submits the SQL statement. After that, the streaming table and its pipeline are managed by Databricks. If you still see a table disappear, the table's pipeline event log and the workspace audit log should show what deleted it. It would also help to know your driver version, whether the table is in Unity Catalog or the Hive metastore, and the full
CREATEstatement you run.vuanhphung commentedon Nov 13, 2026
Please reopen this with more details if you are still running issues.