Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

make outputs error #16204

Merged
merged 10 commits into from
Oct 2, 2017
Merged

make outputs error #16204

merged 10 commits into from
Oct 2, 2017

Conversation

jbardin
Copy link
Member

@jbardin jbardin commented Sep 27, 2017

Allow outputs to generate errors.

Since this can break existing configuration relying on outputs failing silently, the output failure is blocked by the TF_OUTPUT_ERRORS feature flag.

A lot of module "bugs" can be traced back to an output that failed to generate a value. An interpolation error in an output should be handled the same as anywhere else, so that evaluation doesn't continue with possibly invalid data.

Start by making output interpolation errors fatal.
We bypass the known unavoidable errors in 3 ways:

  • During Input the errors are logged, the value is set to unknown, and the walk continues.

  • During Destroy the outputs themselves are destroyed, by ignoring their value and removing them from the state.

  • During a targeted walk, we prune outputs that can't resolve all their dependencies as long as nothing else targeted depends on the output.

  • The final case where output would fail was when applying a destroy plan, which would be treated as an apply. This is fixed by marking the plan for "Destroy", then transferring that to the context so that walkDestroy is used rather than walkApply.

When an output value is needed during a destroy walk, the destruction of the output can cause dependent resources to fail to destroy. This also applies to local values. This is fixed by creating the DestroyValueReferenceTransformer, which walk the graph and reverses the edges for output and local values, so that they are not removed until after their dependents.

@jbardin jbardin force-pushed the jbardin/outputs branch 4 times, most recently from c2844b7 to 7307335 Compare September 29, 2017 19:20
@jbardin jbardin changed the title [WIP] make outputs error make outputs error Sep 29, 2017
func init() {
featureOutputErrors = os.Getenv("TF_OUTPUT_ERRORS") != ""
if featureOutputErrors {
fmt.Println("WARNING: output errors feature enabled")
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I remember we've had problems before with early output interfering with... something. Perhaps it was the plugin RPC protocol? If things seem to be working okay then this is probably not a big deal since this is off by default anyway, but just wanted to flag it since I can't remember what exactly the problem was when we did something like this before.

Copy link
Member Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Yeah, I wanted this to be kind of loud, but I didn't think about it getting built into a provider.
This is internal only anyway, so it's probably better to drop it for now.

Ouptuts don't need to be re-evaluated during destroy, since everything
is already in the state, so we can simply remove them.
Remove the Input flag threaded through the input graph creation process
to prevent interpolation failures on module variables.
Use an EvalOpFilter instead to inset the correct EvalNode during
walkInput. Remove the EvalTryInterpolate type, and use the same
ContinueOnErr flag as the output node for consistency and to try and
keep the number possible eval node types down.
Module outputs may not have complete information during Input, because
it happens before refresh. Continue process on output interpolation
errors during the Input walk.
A Targeted graph may include outputs that were transitively included,
but if they are missing any dependencies they will fail to interpolate
later on.

Prune any outputs in the TargetsTransformer that have missing
dependencies, and are not depended on by any resource. This will
maintain the existing behavior of outputs failing silently ni most
cases, but allow errors to be surfaced where the output value is
required.
When working on an existing plan, the context always used walkApply,
even if the plan was for a full destroy. Mark in the plan if it was
icreated for a destroy, and transfer that to the context when reading
the plan.
DestroyValueReferenceTransformer is used during destroy to reverse the
edges for output and local values. Because destruction is going to
remove these from the state, nodes that depend on their value need to be
visited first.
We're going to start merging breaking functgionality behind feature
flags, to reduce the need for long-lived feature branches.
Add some tests for output errors and catch the errors behind the output
errors feature flag.
This test was set to fail once this issue was fixed, and now it's fixed.
@ghost
Copy link

ghost commented Apr 7, 2020

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.

@ghost ghost locked and limited conversation to collaborators Apr 7, 2020
Sign up for free to subscribe to this conversation on GitHub. Already have an account? Sign in.
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

2 participants