Forum Discussion
Linked Dataflows, Tables and Refreshes - confused
I have Dataflow 1 which contains the datasets which refresh hourly, one of the tables is called Actions. I guess you could call this a staging dataflow.
In a second dataflow (Dataflow 2), I want to transform the Actions table significantly for a specific purpose (I want to leave the dataflow 1 Actions table as is) and so it's added it as a "linked table". But then I can't transform a linked table so I need to create a new computed table with a reference to the linked table -- is this only possible in Desktop as I can't see how to do it within Dataflow 2?
Also, will the linked table automatically refresh at the same time as the original in Dataflow 1? (although I assume the computed table will need a separate refresh).
Hope these 2 questions make sense. Thanks.
HI royces99 ,
If you need to have modifications to a dataflow you can do that in Dataflow 2. You link in the Dataflow1 results and then reference (or merge or append) the Dataflow1 result to make a new query with your modifications.
As you mentioned though, the catch is the timing. Dataflow2 does NOT make dataflow results refresh. This is two separate refreshes and so you have to know about how long a normal refresh for Dataflow1 takes to complete and then schedule Dataflow2 to be after that completes. And, I strongly recommend that you leave buffer time - sometimes dataflows do not start at the exact moment you want and sometimes they run longer than you think.
2 Replies
- collinq
Super User
HI royces99 ,
If you need to have modifications to a dataflow you can do that in Dataflow 2. You link in the Dataflow1 results and then reference (or merge or append) the Dataflow1 result to make a new query with your modifications.
As you mentioned though, the catch is the timing. Dataflow2 does NOT make dataflow results refresh. This is two separate refreshes and so you have to know about how long a normal refresh for Dataflow1 takes to complete and then schedule Dataflow2 to be after that completes. And, I strongly recommend that you leave buffer time - sometimes dataflows do not start at the exact moment you want and sometimes they run longer than you think.On my company's IT estate, I can set the Dataflow and Semantic models to refresh on the hour and at half past each hour (so every 30 minutes). Each one only takes a few minutes though. Dos this mean that if Dataflow refreshes at 1pm, then Dataflow 2 won't pick up those updates until 1.30pm and then the semantic model at 2pm? Or will Dataflow 2 update immediately following Dataflow 1 at 1pm (or 1.05pm say)? Just trying to work out how long it takes from a data change at source to make it to the Power BI report for the user, and if eliminating a Dataflow would make it quicker?
