Bug Report: Provider sends UNSPECIFIED enum values (0) on cluster updates

[jean-haley]

Bug Report: Provider sends UNSPECIFIED enum values (0) on cluster updates

PR 150 Created a PR for this fix, as it seems straightforward enough. Please reject if I am incorrect.

Description

The terraform-provider-qdrant-cloud sends gpu_type: 0 and restart_policy: 0 (UNSPECIFIED enum values) during cluster update operations, which the Qdrant Cloud API rejects with a validation error.

Environment

• Provider version: v1.18.0
• Pulumi bridged provider: pulumi_qdrant_cloud==1.18.0

Error

error: Invalid argument for cluster update [cluster-id]: invalid request: validation error:
 - cluster.configuration.gpu_type: value must not be in list [0] (value: 0) [enum.not_in]
 - cluster.configuration.restart_policy: value must not be in list [0] (value: 0) [enum.not_in]

Steps to Reproduce

1. Create a cluster without specifying gpu_type or restart_policy
2. Run terraform apply (or pulumi up) again — even with no changes
3. Provider attempts an update and sends gpu_type: 0, restart_policy: 0
4. API rejects the request

Root Cause

PR #147 fixed the flatten (read) path to skip storing UNSPECIFIED values in state. However, the expand (write) path in expandClusterConfiguration() still accepts UNSPECIFIED as a valid enum:

// schemas_accounts_clusters.go:972-976
if v, ok := item[dbConfigGpuTypeFieldName]; ok {
    gt, gtOK := qcCluster.ClusterConfigurationGpuType_value[v.(string)]
    if gtOK {  // ← "CLUSTER_CONFIGURATION_GPU_TYPE_UNSPECIFIED" passes this check
        config.GpuType = newPointer(qcCluster.ClusterConfigurationGpuType(gt))
    }
}

Since ClusterConfigurationGpuType_value includes "CLUSTER_CONFIGURATION_GPU_TYPE_UNSPECIFIED": 0, the expand function sets GpuType to a pointer to 0 instead of leaving it nil.

Possible Fix (if helpful)

One approach might be to filter out UNSPECIFIED values in the expand path, similar to how PR #147 filters them in the flatten path:

if v, ok := item[dbConfigGpuTypeFieldName]; ok {
    gt, gtOK := qcCluster.ClusterConfigurationGpuType_value[v.(string)]
    if gtOK && gt != int32(qcCluster.ClusterConfigurationGpuType_CLUSTER_CONFIGURATION_GPU_TYPE_UNSPECIFIED) {
        config.GpuType = newPointer(qcCluster.ClusterConfigurationGpuType(gt))
    }
}

The same pattern would apply to restart_policy and rebalance_strategy.

Submitted a PR that I believe will work to fix this.

Additional Context

• The update is triggered even when Pulumi/Terraform detects no diff ([diff: ] is empty)
• Using ignore_changes on configuration does not prevent the issue
• Clusters without GPU have no valid value to send — the field must be omitted entirely