core: update destroy graph pruning to handle ephemeral resources #35890
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
With the missing destroy node types sorted out, this transformer can now be designed around finding the nodes we need to keep for destroy operations, rather than the whack-a-mole game of detecting nodes we don't want to keep. When executing a destroy operation, we can never be certain that the state will contain all the data required to evaluate the entire configuration. This could be because of a previously failed apply operation, or external changes made out of band. This means that once we process the configuration to get the providers hooked up, we need to remove any nodes that correspond to configuration values which may not be available to evaluate. In a destroy operation we primarily need three categories of nodes to complete the plan and apply: - Nodes corresponding to values in the state - Providers required to destroy those values obtained from the state - Any node which a provider depends on, to ensure its configuration can be fully evaluated Other than that minimal set of nodes, the rest are no longer necessary and should be skipped to avoid errors which could block the apply operation.
DanielMSchmidt
approved these changes
Oct 24, 2024
|
Reminder for the merging maintainer: if this is a user-visible change, please update the changelog on the appropriate release branch. |
|
I'm going to lock this pull request because it has been closed for 30 days ⏳. This helps our maintainers find and focus on the active contributions. |
Sign up for free
to subscribe to this conversation on GitHub.
Already have an account?
Sign in.
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Ephemeral resources were the straw that broke the
pruneUnusedNodesTransformer's back. This transformer has always been in a tough spot, trying to maintain legacy behaviors around destroy, while keeping destroy operations from failing entirely from configuration evaluation errors. As we've tidied up graph creation over many prior versions, there were just a few node types which were incorrectly classified that were preventing us from inverting thepruneUnusedNodesTransformerlogic to the more reasonable task of finding what to keep, rather than finding what to exclude.With the missing destroy node types sorted out, the
pruneUnusedNodesTransformerno longer needs to keep up with the whack-a-mole game of detecting nodes which might need to be skipped. When executing a destroy operation, we can never be certain that the state will contain all the data required to evaluate the entire configuration. This could be because of a previously failed apply operation, or external changes made out of band. This means that once we process the configuration to get the providers hooked up, we need to somehow remove any nodes that correspond to configuration values that may not be available to evaluate.In a destroy operation we primarily need three categories of nodes to complete the plan and apply:
Other than that minimal set of required nodes, the rest are no longer necessary and should be skipped to avoid any errors that could block the apply operation.