What are the best steps to troubleshoot a dataflow that has stopped?

When I'm troubleshooting a dataflow, my first step is to go to the details page for the dataflow. To get there, go to the Data Center and click the wrench menu for the dataflow you want to troubleshoot. Choose the "Details" option.

From the details page, you can see the history of the attempted runs of the dataflow. If it has failed you'll see the Start Time, End Time, Duration, and then "Failed" as the Status. If you hover over the Failed message, the error message that was logged when it failed. Most of the time, this is an error that came back directly from the database and can seem fairly intimidating. In that error, though, there's generally information to glean about what went wrong (missing column, incompatible data types, etc.). If it isn't clear from the error message alone, click on the "Failed" message and you'll see each transform in the dataflow with a green or red box around it. The green ones succeeded and the red one failed. Hopefully with the combination of the error message and the transform that failed, you can track down the error.

As a note, it has been my experience that if a dataflow was working and then suddenly stopped, it very frequently comes down to having a change in the schema of one of the input datasets. It can stop unions for working if the number or types of the columns changes. It call also break transforms if fields disappear that were being used in the transform.

I hope these ideas help!