Upgrade Failed - connection references

[sunilmhatugade]

Does this bug already exist in our backlog?

• I have checked and confirm this is a new bug.

Describe the issue

We are trying to upgrade COE Kit from 4.17 to 4.43 (sept month release). While upgrading we are facing issue.

Kindly Assist

Also please let us know with fix this when we release will come.
we have to implement for client also so Which one is stable release for upgradation.

Expected Behavior

No response

What solution are you experiencing the issue with?

Core

What solution version are you using?

4.17

What app or flow are you having the issue with?

Core component

What method are you using to get inventory and telemetry?

None

Steps To Reproduce

No response

Anything else?

No response

[Jenefer-Monroe]

Unfortunately it looks like you've hit a product bug around upgrade. A few other people have hit this too. Please see #8466.

The connection

This is likely the legacy Dataverse connection that is at play, why it wont tell you the name I don’t know.
But this is a connection that we have removed from the kit and that is causing issues in the product’s upgrade code.

The likely issue

The issue is around dependency ordering.
Flow X uses Connection Y
Both Flow X and Connection Y are deleted as part of the upgrade
The product is attempting to delete Connection Y before Flow X, which is cannot do.

Find impacted flows

Please go to the default solution > connection references > look at the dependencies for both of these connection references

How to work around the bug

As a result, in order to upgrade you will break the dependencies.
You will need to:

1. Delete the _Upgrade solution (if one was left behind)
2. Open all the flows above, delete most of whats in them, and resave
3. Do this until the connection reference no longer shows dependencies
4. Try the upgrade again

Step 2 example

Example of how to do the workaround for one of the flows:

Thats just awful!

Yes I apologize for the hassle.
The product team is aware of this issue but the fix has not yet come out.
And another user has succeeded with this workaround. See #8466