Forum Discussion
Kudo CommanderDates dataflow suddenly reverting to earlier version, dropping fields
I received a data refresh error on a PBIX in the service this week that failed due to a missing column in my Dates dataflow. When I opened the report to investigate, all the fields I expected to see were present in the fields list of the Dates table, but when I refreshed that data source in the PBIX, a number of the fields I'd created just disappeared. I then went to the dataflow in the service to confirm that the code for my additional fields there was indeed missing, and it was.
I'm wondering if anyone else has seen/is seeing this?
Thanks for the reply, edhans . I guess that's possible. Here's the confounding part: The report in question didn't fail in the couple days after I'd last done development on the Dates dataflow, which I'd expect to have happened if I somehow managed to revert to a cached version of my dataflow. I guess this is a good reminder to save the code for these tables somewhere outside the service. Mea culpa.
6 Replies
- edhans
Community Champion
I've never seen a dataflow revert, but what could have happened is if you opened a dataflow and got a message similar to "this dataflow wasn't closed properly would you like to recover from the cache" and you do, or don't, it can change what is there. Unfortunately, it will not tell you what pressing Yes or No will actually do to the code, so it is a shot in the dark. Unless I had a browser crash, I almost always hit no. That browser tab instance could have a cache from a few days ago.
MarkPalmbergThanks for the reply, edhans . I guess that's possible. Here's the confounding part: The report in question didn't fail in the couple days after I'd last done development on the Dates dataflow, which I'd expect to have happened if I somehow managed to revert to a cached version of my dataflow. I guess this is a good reminder to save the code for these tables somewhere outside the service. Mea culpa.
- edhans
You can export the JSON and restore the dataflow, but it will assign the dataflow a new ID, which means you'll need to tweak that code in your dataset. Kind of a pain... but all M code for the entire dataflow will be in one file.
