The 03:30 JST pipeline broke again. State backup looks clean, but Terraform 1.9.5 paired with nice-cxone 1.16.3 throws a 409 Conflict whenever the nice_cxone_data_action resource tries to bind to an existing HTTP integration. Console shows the integration is active. API calls to /api/v2/integrations return a 200. The Terraform apply hangs for twelve minutes before dropping the error.
Tried forcing recreation with lifecycle ignore_changes on the integration_id field. Didn’t help. Provider logs show the PUT request hitting /api/v2/data-actions with a malformed payload. The schema_version parameter seems to drop out during serialization. Rolled back to provider 1.15.8 just to verify. That version applies cleanly, but the data action stays disabled in the console.
Environment runs on a self-hosted runner in Tokyo. AWS credentials are scoped correctly. The integration itself points to a VPC endpoint for Lambda. Nothing changed on the backend side. Checked the audit trail. No concurrent writes. State drift backup confirms the remote state matches the local plan.
The provider documentation mentions a retry mechanism for 409s, but it’s doing jack all. Console stays quiet. Terraform just keeps looping the apply step. The HTTP integration uses a custom header for auth. Maybe the provider is stripping it out during the binding phase.
Here is the exact payload the provider sends before the crash:
{
"name": "txn-validation-step",
"integration_id": "a8f3-22b1-cxone-99z",
"schema_version": "1.0.0",
"timeout_ms": 3000,
"enabled": true
}