Forum Discussion
Deployment Pipelines - Dataset is different after a "Successful" Promotion from DEV to TEST
Hello,
We have been using Deployment Pipelines fairly successfully for 2 months now and working through issues. We need some ideas on how to best troubleshoot what is causing the dataset to be different, after a successful deployment from DEV to TEST.
What we did:
1. Downloaded Power BI Desktop Oct 2020 version
2. Implemented RLS into our dataset
3. Published "Sales" PBIX file into Deployment Pipeline DEV workspace
4. Had an administrator "Take Over" dataset ownership from publisher developer, made sure credentials to our 1 data source was valid and refreshed dataset successfully
5. In Deployment Pipelines interface, Administrator selected "Sales" report and dataset from DEV and promoted to TEST workspace "Successfully", however after deployment, the dataset was not equivalent, but the report was.
6. While troubleshooting, we had our BI "Sales" report developer rename the "Sales" report to "Sales_Test" and do the deployment of the report and dataset together from DEV to TEST, and the dataset was equivalent.
We dont want the Developer doing the deployments, so do you have any troubleshooting ideas on why things are different?
We are not using any of the following Unsupported items:
Unsupported items
Deployment pipelines doesn't support the following items:
Datasets that do not originate from a .pbix
Reports based on unsupported datasets
Paginated reports
Dataflows
PUSH datasets
Workbooks
41 Replies
Hello I have exactly the same issue now.
After deployment process from dev to test piepline shows the datasets are different. What's more if the dataset is published directly to the PROD then after deployment from Test to PROD we have 2 datasets with the same name.
I've already remove all datasets from DEV and TEST workspaces and try to deloy everything once again, but it still not working.
Any idea what can I do?
Data-RainerAdvocate IV
It has been two years since this thread was posted and I am experiencing the exact same issue with our deployment pipeline. We have successfully operated our dataset promotion for over a year without any issue ever. Suddenly since last week our TEST & PROD pipeline show our main dataset as "Different", even though they are identical as they always have been and have been successfully deployed from DEV to TEST to PROD.
Nimrod_Shalit
Is there any progress on identifying possible causes?- LeeBenjamin
Microsoft Employee
Hi Data-Rainer ,
Status of "Different" for a deployed item shouldn't be expected right after a successful deployment (could be though after a failed deployment). Therefore, it could be one of the two:
- A bug - Please open a ticket and provide the deployment id (can be copied from the deployment's record on 'Deployment History')
- A change was done next to the deployment completion. The dataset could be changed by several users on several places (not just through pipeline) so it could be someone else made a change.
In general, there could be several reasons for deployment pipeline to tag the "Different" label to an item, which can be ruoghly categorized as follow:
- Setting change: renaming an item, changing an item's sensitivity label.
- Code change: which means content/schema changes.
- Deployment rule was set and deployment is required to apply it - The rule will be considered for the compare action, but since deployment wasn't executed yet, a difference will be identified for the item.
we're working these days on adding indication for 'Different' items, saying which of the above two reasoning categories is the reasone for the DIFF label. moreover, in case item got indication of code change, it would be possible for users to review the code changes (Read-only. Could be for only some item types at the begining).
Best regards,
Lee
Deployment pipeline PM
Exact same issue since this week, here too.
A few tests that I have done :
* Publish the dataset with a diferent name to force te creation of a new object, push to test --> says the files are different.
* I recover versions from setembre and octobre that worked fine at the time, push to test --> keeps saying the are different.
* I started removing content from the dataset to check if it is any part of the dataset that is broken and even with a simple single table connected to the source it keeps saying the files are different.
* When I test a new file with a dummy manual source it works fine.
It seems to me a bug on the deployment pipelines ?
- Nimrod_Shalit
Power BI Team
Few questions to better understand:
1. did you configure rules on the 'Sales' dataset? if so, were they already applied?
2. when you said 'different', did you mean that they were labeled as 'different', or you actually spotted a difference between the datasets?
1. The BI developer applied the Role Level Security rules to the dataset before publishing to DEV workspace.
2. The deployment pipelines interface shows the "sales" dataset as non-equivalent between DEV and TEST, but I cannot spot a differences when comparing files
karoAdvocate V
Hi,
Anonymous do you have any updates? It seems that I am experiencing similar issue.
Nimrod_Shalit any known issues from Microsoft regarding this case?
Regards,
Karo
Hello,
I stopped to have this strange behavior after separating dataset and report in two different .pbix files as recommended in the good practices of deployment pipelines:
https://docs.microsoft.com/en-us/power-bi/create-reports/deployment-pipelines-best-practices#separate-modeling-development-from-report-and-dashboard-development
To split it, check the following link :
https://www.havensconsulting.net/blog-and-media/splitting-a-power-bi-pbix-file-into-a-dataset-and-lean-report-file
With that, the changes done in the report or in the dataset become more independent what is better for deployment.
I hope it can help,
Regards,
- KNP
Super User
I've had a similar issue and after a bit of testing, I've found a workaround for my scenario. Posting here in case it helps others.
The change that broke it
My dataset is separate from my report.
Deployment pipelines were previously working without issue.
The only change I made was to rename the only security role in the dataset and deploy to dev.
Confirming the issue
To check that my assumption was correct, I changed the name of the security role back to what it was and redeployed to dev. Deployment pipelines worked again.
The workaround
I needed the security role renamed so to try and get dev and test to align, I deployed my PBIX with the renamed security role back to dev and then also manually deployed my PBIX directly to test. No need to delete first.
Pipelines now works again.
Like I say, might not be suitable for all cases, but hopefully it is helpful to some.
