We tried the downgrade and it didnt work for us either.
After some investigation we found that refreshes in PBI online through the spark connector (to databricks) fails when the data contains datetime and/or decimal types. The same dataset would refresh against the same cluster in powerbi desktop however.
Within our databricks jobs we converted datetime to unix_time and decimal to float or double and the refresh works in both desktop and online PBI versions.
Hope this helps some of you out there.
We are currently running Databricks 6.3, which should be fine, and it fails with the same error message. We opened a ticket with Microsoft.
I just experienced this issue. Thanks for the discussion and possible workarounds as well as the message from the Microsoft support page:
"Power BI customers using Databricks as a Datasource may encounter the following message during refresh: "Conversion from number to string failed due to undersized character buffer". Engineers have identified the root cause as an incompatibility between Databricks version 6.4 (and above) and connector used by Power BI service. Customers can avoid this issue by using a Databricks version lower than 6.4 until the fix is deployed. ETA for a fix will be provided soon. "
Same here, tried 6.3 as stated by Microsoft on the portal, but only works on 5.5 LTS (also suggested by Microsoft in response to our case and some of you guys)
Check out a full recap of the month!
Check out the winners of the recent 'Can You Solve These?' community challenge!
Join us for an in-depth look at the new Power BI features and capabilities at the free Microsoft Business Applications Launch Event.