graph refactor causes regression in plan

[pmoust]

After #1010 with a state built with 0.3.7 some nodes on terraform plan are being removed from state and added back in the execution plan, even though the state before plan is fine and matches the actual cloud state.
Reverting our terraform build pre #1010 .

I tried to narrow it down to a minified reproducible case, but no luck yet.

What would be optimal way to trace down the graph plan pre and after #1010 TF builds in order to compare?
Setting DEBUG display would suffice?

[catsby]

Probably related to / fixed by #1046

[pmoust]

Indeed #1046 did it. Thanks guys.

[catsby]

Great to hear, thank you for reporting the issue!

[pmoust]

Hmm, seems we are experiencing this again even after #1046

[pmoust]

Mentioned in PeoplePerHour#11 this is almost the same output as when first opened the ticket after #1010 . Now also an edge node (that is not dependant on the other changes shown in plan) is marked for record change.
And again I fail in creating a small reproducible case. The issue does exist though,and is not present in our current head.

[mitchellh]

Let me know if you can get more luck in a repro case. I'll keep thinking about it as well.

[pmoust]

TF_LOG @ PeoplePerHour#11 (comment)

[mitchellh]

@pmoust Found it, and made a PR to fix here #1086. Note that this will cause your diff to delete two resources. If you look, these resources always existed in the state, but were always ignored. I imagine these are just orphan resources in your infrastructure that you forgot were running since TF never "saw" them, so the diff should be correct.

Whew, a doozy.

[ghost]

I'm going to lock this issue because it has been closed for 30 days ⏳. This helps our maintainers find and focus on the active issues.

If you have found a problem that seems similar to this, please open a new issue and complete the issue template so we can capture all the details necessary to investigate further.