fix(monitor-alert): handle legacy alerts with metricId fallback

Legacy metric alerts (pre-2024 schema) store the metric identifier in
the flat `metricId` field instead of the nested `metric.id` object.
When importing these alerts, the provider now falls back to `metricId`
if `metric.id` is empty, making them manageable via Terraform.

Also adds StringIsNotEmpty validation on the metric field to surface
a clear error at plan time instead of a cryptic API 422.

Closes #708

Author: tembleking

sysdig/internal/client/v2/model.go

@@ -773,6 +773,7 @@ type AlertV2ConfigMetric struct {
 	GroupAggregation string                  `json:"groupAggregation"`
 	TimeAggregation  string                  `json:"timeAggregation"`
 	Metric           AlertMetricDescriptorV2 `json:"metric"`
+	MetricID         string                  `json:"metricId,omitempty"` // Legacy API field, used as fallback when Metric.ID is empty
 	NoDataBehaviour  string                  `json:"noDataBehaviour"`
 
 	Range    int `json:"range"`

sysdig/internal/client/v2/model_test.go

@@ -0,0 +1,84 @@
+package v2
+
+import (
+	"encoding/json"
+	"testing"
+)
+
+func TestAlertV2MetricUnmarshalLegacyMetricID(t *testing.T) {
+	// Legacy API response: has "metricId" but no "metric.id"
+	legacyJSON := `{
+		"alert": {
+			"id": 1553848,
+			"name": "MysqlMateo",
+			"type": "MANUAL",
+			"config": {
+				"metricId": "mysql_global_status_threads_connected",
+				"groupAggregation": "avg",
+				"timeAggregation": "avg",
+				"threshold": 125.0,
+				"conditionOperator": ">",
+				"noDataBehaviour": "DO_NOTHING",
+				"range": 60,
+				"duration": 0
+			}
+		}
+	}`
+
+	var wrapper alertV2MetricWrapper
+	err := json.Unmarshal([]byte(legacyJSON), &wrapper)
+	if err != nil {
+		t.Fatalf("failed to unmarshal legacy alert JSON: %v", err)
+	}
+
+	alert := wrapper.Alert
+
+	if alert.Config.MetricID != "mysql_global_status_threads_connected" {
+		t.Errorf("expected MetricID to be %q, got %q", "mysql_global_status_threads_connected", alert.Config.MetricID)
+	}
+
+	// metric.id should be empty since the legacy response doesn't include it
+	if alert.Config.Metric.ID != "" {
+		t.Errorf("expected Metric.ID to be empty for legacy alert, got %q", alert.Config.Metric.ID)
+	}
+}
+
+func TestAlertV2MetricUnmarshalCurrentMetricID(t *testing.T) {
+	// Current API response: has both "metric.id" and "metricId"
+	currentJSON := `{
+		"alert": {
+			"id": 1760356,
+			"name": "cpu usage spike alert",
+			"type": "MANUAL",
+			"config": {
+				"metric": {
+					"id": "sysdig_container_cpu_quota_used_percent"
+				},
+				"metricId": "sysdig_container_cpu_quota_used_percent",
+				"groupAggregation": "sum",
+				"timeAggregation": "avg",
+				"threshold": 90.0,
+				"conditionOperator": ">",
+				"noDataBehaviour": "DO_NOTHING",
+				"range": 60,
+				"duration": 0
+			}
+		}
+	}`
+
+	var wrapper alertV2MetricWrapper
+	err := json.Unmarshal([]byte(currentJSON), &wrapper)
+	if err != nil {
+		t.Fatalf("failed to unmarshal current alert JSON: %v", err)
+	}
+
+	alert := wrapper.Alert
+
+	if alert.Config.Metric.ID != "sysdig_container_cpu_quota_used_percent" {
+		t.Errorf("expected Metric.ID to be %q, got %q", "sysdig_container_cpu_quota_used_percent", alert.Config.Metric.ID)
+	}
+
+	if alert.Config.MetricID != "sysdig_container_cpu_quota_used_percent" {
+		t.Errorf("expected MetricID to be %q, got %q", "sysdig_container_cpu_quota_used_percent", alert.Config.MetricID)
+	}
+}

sysdig/resource_sysdig_monitor_alert_v2_metric.go

@@ -275,7 +275,11 @@ func updateAlertV2MetricState(d *schema.ResourceData, alert *v2.AlertV2Metric) e
 
 	_ = d.Set("group_aggregation", alert.Config.GroupAggregation)
 
-	_ = d.Set("metric", alert.Config.Metric.ID)
+	metricID := alert.Config.Metric.ID
+	if metricID == "" {
+		metricID = alert.Config.MetricID
+	}
+	_ = d.Set("metric", metricID)
 
 	_ = d.Set("no_data_behaviour", alert.Config.NoDataBehaviour)