diff --git a/.changelog/2949.txt b/.changelog/2949.txt
new file mode 100644
index 0000000000..0d3c22ae9e
--- /dev/null
+++ b/.changelog/2949.txt
@@ -0,0 +1,3 @@
+```release-note:enhancement
+resource/tencentcloud_kubernetes_cluster: support set `instance_delete_mode`
+```
\ No newline at end of file
diff --git a/tencentcloud/services/tke/resource_tc_kubernetes_cluster.go b/tencentcloud/services/tke/resource_tc_kubernetes_cluster.go
index 55cb807ddc..054f958e8d 100644
--- a/tencentcloud/services/tke/resource_tc_kubernetes_cluster.go
+++ b/tencentcloud/services/tke/resource_tc_kubernetes_cluster.go
@@ -1325,6 +1325,12 @@ func ResourceTencentCloudKubernetesCluster() *schema.Resource {
 				Optional:    true,
 				Description: "CDC ID.",
 			},
+
+			"instance_delete_mode": {
+				Type:        schema.TypeString,
+				Optional:    true,
+				Description: "The strategy for deleting cluster instances: terminate (destroy instances, only support pay as you go cloud host instances) retain (remove only, keep instances), Default is terminate.",
+			},
 		},
 	}
 }
@@ -1945,6 +1951,10 @@ func resourceTencentCloudKubernetesClusterDelete(d *schema.ResourceData, meta in
 	request.ClusterId = helper.String(clusterId)
 
 	instanceDeleteMode := "terminate"
+	if v, ok := d.GetOk("instance_delete_mode"); ok {
+		instanceDeleteMode = v.(string)
+	}
+
 	request.InstanceDeleteMode = &instanceDeleteMode
 
 	if v, ok := d.GetOk("resource_delete_options"); ok {
diff --git a/tencentcloud/services/tke/resource_tc_kubernetes_cluster.md b/tencentcloud/services/tke/resource_tc_kubernetes_cluster.md
index d73252dd17..d6e43ee1e1 100644
--- a/tencentcloud/services/tke/resource_tc_kubernetes_cluster.md
+++ b/tencentcloud/services/tke/resource_tc_kubernetes_cluster.md
@@ -5,6 +5,8 @@ Provide a resource to create a kubernetes cluster.
 ~> **NOTE:** We recommend this usage that uses the `tencentcloud_kubernetes_cluster` resource to create a cluster without any `worker_config`, then adds nodes by the `tencentcloud_kubernetes_node_pool` resource.
 It's more flexible than managing worker config directly with `tencentcloud_kubernetes_cluster`, `tencentcloud_kubernetes_scale_worker`, or existing node management of `tencentcloud_kubernetes_attachment`. The reason is that `worker_config` is unchangeable and may cause the whole cluster resource to `ForceNew`.
 
+~> **NOTE:** Executing `terraform destroy` to destroy the resource will default to deleting the node resource, If it is necessary to preserve node instance resources, Please set `instance_delete_mode` to `retain`.
+
 Example Usage
 
 Create a basic cluster with two worker nodes
@@ -397,7 +399,6 @@ resource "tencentcloud_kubernetes_cluster_endpoint" "example" {
     tencentcloud_kubernetes_node_pool.example
   ]
 }
-
 ````
 
 Use Kubelet
@@ -699,6 +700,7 @@ resource "tencentcloud_kubernetes_cluster" "test_node_pool_global_config" {
 ```
 
 Using VPC-CNI network type
+
 ```hcl
 variable "availability_zone" {
   default = "ap-guangzhou-1"
@@ -756,6 +758,7 @@ resource "tencentcloud_kubernetes_cluster" "managed_cluster" {
 ```
 
 Using ops options
+
 ```
 resource "tencentcloud_kubernetes_cluster" "managed_cluster" {
   # ...your basic fields
@@ -868,5 +871,5 @@ Import
 tke cluster can be imported, e.g.
 
 ```
-$ terraform import tencentcloud_kubernetes_cluster.test cls-xxx
+$ terraform import tencentcloud_kubernetes_cluster.example cls-n2h4jbtk
 ```
\ No newline at end of file
diff --git a/website/docs/r/kubernetes_cluster.html.markdown b/website/docs/r/kubernetes_cluster.html.markdown
index 5da4e8eba4..07edbc6348 100644
--- a/website/docs/r/kubernetes_cluster.html.markdown
+++ b/website/docs/r/kubernetes_cluster.html.markdown
@@ -16,6 +16,8 @@ Provide a resource to create a kubernetes cluster.
 ~> **NOTE:** We recommend this usage that uses the `tencentcloud_kubernetes_cluster` resource to create a cluster without any `worker_config`, then adds nodes by the `tencentcloud_kubernetes_node_pool` resource.
 It's more flexible than managing worker config directly with `tencentcloud_kubernetes_cluster`, `tencentcloud_kubernetes_scale_worker`, or existing node management of `tencentcloud_kubernetes_attachment`. The reason is that `worker_config` is unchangeable and may cause the whole cluster resource to `ForceNew`.
 
+~> **NOTE:** Executing `terraform destroy` to destroy the resource will default to deleting the node resource, If it is necessary to preserve node instance resources, Please set `instance_delete_mode` to `retain`.
+
 ## Example Usage
 
 ### Create a basic cluster with two worker nodes
@@ -917,6 +919,7 @@ The following arguments are supported:
 * `globe_desired_pod_num` - (Optional, Int, ForceNew) Indicate to set desired pod number in node. valid when enable_customized_pod_cidr=true, and it takes effect for all nodes.
 * `ignore_cluster_cidr_conflict` - (Optional, Bool, ForceNew) Indicates whether to ignore the cluster cidr conflict error. Default is false.
 * `ignore_service_cidr_conflict` - (Optional, Bool, ForceNew) Indicates whether to ignore the service cidr conflict error. Only valid in `VPC-CNI` mode.
+* `instance_delete_mode` - (Optional, String) The strategy for deleting cluster instances: terminate (destroy instances, only support pay as you go cloud host instances) retain (remove only, keep instances), Default is terminate.
 * `is_non_static_ip_mode` - (Optional, Bool, ForceNew) Indicates whether non-static ip mode is enabled. Default is false.
 * `kube_proxy_mode` - (Optional, String) Cluster kube-proxy mode, the available values include: 'kube-proxy-bpf'. Default is not set.When set to kube-proxy-bpf, cluster version greater than 1.14 and with Tencent Linux 2.4 is required.
 * `labels` - (Optional, Map, ForceNew) Labels of tke cluster nodes.
@@ -1116,6 +1119,6 @@ In addition to all arguments above, the following attributes are exported:
 tke cluster can be imported, e.g.
 
 ```
-$ terraform import tencentcloud_kubernetes_cluster.test cls-xxx
+$ terraform import tencentcloud_kubernetes_cluster.example cls-n2h4jbtk
 ```