The 05:30 JST deployment pipeline’s doing jack all right now. Terraform 1.9.8 paired with nice-cxone 1.20.5 throws a 422 Unprocessable Entity when pushing updates to the trigger_condition_serialization block. Cleared the local cache and ran a fresh state pull. Console just spits out the same validation error about mismatched enum values for interaction_type.
Tried downgrading to provider 1.20.3. Same result. Switched to raw REST calls via Postman to bypass the Terraform wrapper entirely. API accepts the payload without hesitation. It’s pointing straight to the provider’s serialization logic stripping the nested media_capture_scope array. State drift backup shows zero uncommitted changes.
Ran terraform plan -detailed-exitcode to catch hidden diffs. Exit code 2. The diff highlights a phantom modification on the recording_retention_days attribute even though the HCL matches the live environment exactly. Forced a terraform state rm on the policy resource and re-imported. Pipeline choked on the exact same 422 response during the apply phase.
Logs show the request hits PATCH /api/v2/quality/screen-recording/policies/{policyId} with a truncated body. Provider seems to be dropping the agent_group_id reference during the merge step. You’ll notice the timeout threshold keeps hitting 30 seconds before the handshake drops. You can’t skip the validation layer without triggering the same enum mismatch.
{
"error_code": "VALIDATION_ERROR",
"message": "Field trigger_condition_serialization contains invalid enum value",
"invalid_values": ["voice_and_screen", "screen_only"]
}
Provider 1.20.5 definitely mangles the nested object mapping on screen recording policies. Left the debug trace running while the CI runner spins.