We provide an OData4 producer that offers customers access to a wide range of platforms such as AFAS Profit, Twinfield, Teamleader, Chargebee or Exact Online. Part of the user community uses Power BI desktop, but several use dataflows.
We find that at last during the last 6 months the use of dataflows is unreliable. Downloads larger than 10 MB (characters, uncompressed, larger than 1 MB compressed ) end in an error:
OData: Unable to read data from the transport connection: An existing connection was forcibly closed
From our monitoring screen on https://bridge-online.cloud/monitoring we see that the data was already returned to the Power BI service approximately 60 seconds before the error occurs. This is sometimes 50 seconds and sometimes 65 seconds, but always around one minute after the OData stream is closed on our side with a HTTP 200 status. Clocks are all synchronized to within 10-15 seconds maximum time difference. We can see that even the TCP ports have already been closed long before the error occurs.
Smaller downloads - even when downloaded in parallel - are succesfull.
It seems that the dataflows service is chewing on the data for 1 minute and then fails somehow with a generic and incorrect error message.
Sofar the workarounds available are chopping up the download in smaller dan 10 MB size parts or reverting to the Power BI desktop.
However, it introduces churn and decreases customer satisfaction so a solution would be helpful.
Complicating is that the error occurs outside our software; we can not detect which customers run into this problem.
What can users of OData4 feeds do to stay on dataflows with larger OData4 downloads?
