Introduce an import heuristic to drop entries that fail validators #2314
Assignees
Labels
Comments
t0yv0
added
kind/enhancement
This was referenced Aug 14, 2024
t0yv0
added
area/import
and removed
needs-triage
corymhall
added a commit
that referenced
this issue
Aug 22, 2024
> This is a reroll of #2277 which was reverted in #2299 > I've added additional handing for secret values During an import operation property values are read from the cloud and turned into inputs which then becomes part of the generated code. For some properties (Computed & Optional) the cloud can return values that fail validation. Since these are computed & optional the provider will allow the user to not specify a value. In these cases the import would succeed, but if the user took the generated code as is and attempted to run pulumi up it would fail due to the validation errors. Rather than generate invalid inputs we will instead run Validate during import and then processes any resulting validation errors. If any properties fail validation (and they match our other cases) then we remove those inputs so that the generated program code will not contain the invalid inputs NOTE: we want to (for now at least) scope this as narrowly as possible because we do not want to create a scenario where the resulting inputs create a different error (e.g. removing an invalid required property). In these cases we will just allow the invalid configuration through and the user will have to decide what to do. fixes pulumi/pulumi-aws#4054, re #2314
|
@corymhall I think #2337 has this mostly done but didn't tackle nested properties, that sounds right? Leaving it in the backlog for now until we have more evidence like linode 373 that this needs fixing. |
|
Yep that is correct. |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Hello!
Issue details
Import for bridged providers may return data from
Read()that does not pass validations. This causes Pulumi to report warnings.Since the issue does not manifest as badly in TF as in Pulumi, the underlying Read() implementations are unlikely to be fixed upstream at scale. The suggestion therefore is to introduce heuristic-like code in the bridge that tries to autocorrect Read() results and drop entries that fail to pass validators. A similar tactic was applied in #1948
Initial implementation in #2337
Todo:
Affected area/feature
Affects import. See also:
The text was updated successfully, but these errors were encountered: