update docs for new plugins

gocrane · Jun 13, 2023 · 94f37f4 · 94f37f4
1 parent ba75922
commit 94f37f4
Show file tree

Hide file tree

Showing 7 changed files with 325 additions and 14 deletions.
diff --git a/site/content/en/docs/Tutorials/Recommendation/pv-recommendation.md b/site/content/en/docs/Tutorials/Recommendation/pv-recommendation.md
@@ -0,0 +1,37 @@
+---
+title: "PV 推荐"
+description: "PV 推荐功能介绍"
+weight: 15
+---
+
+Service 推荐通过扫描集群中 Service 的运行状况，帮助用户找到闲置的 Kubernetes Service。
+
+## 动机
+
+通常在 Kubernetes 中我们会使用 Service + Workload 来自动创建和管理负载均衡并 将负载均衡挂载到应用上，在日常的运营中难免会出现空闲和低利用率的负载均衡，浪费了大量成本，Service 推荐尝试帮助用户找到这部分 Service 来实现成本优化。
+
+## 推荐示例
+
+```yaml
+
+```
+
+在该示例中：
+
+- 推荐的 TargetRef 指向了 PV:
+- 推荐类型为 PV 推荐
+- action 是 Delete，这里只是给出建议
+
+## 实现原理
+
+PV 推荐按以下步骤完成一次推荐过程：
+
+1. 扫描集群中所有 PV，找到 PV 对应的 Pod 列表
+2. 如果 PV 没有对应的 PVC，则判断为闲置 PV
+3. 如果没有 Pod 关联这个 PV 和 PVC，则判断为闲置 PVC
+
+## 参数配置
+
+目前 PV 推荐没有参数配置。
+
+如何更新推荐的配置请参考：[**推荐框架**](/zh-cn/docs/tutorials/recommendation/recommendation-framework)
diff --git a/site/content/en/docs/Tutorials/Recommendation/recommendation-framework.md b/site/content/en/docs/Tutorials/Recommendation/recommendation-framework.md
@@ -128,6 +128,8 @@ Currently, Crane support these Recommenders:
 - [**Replicas Recommendation**](/docs/tutorials/recommendation/replicas-recommendation): Use the HPA algorithm to analyze the actual usage of applications and recommend more appropriate replicas configurations.
 - [**HPA Recommendation**](/docs/tutorials/recommendation/hpa-recommendation): Scan the Workload in a cluster and recommend HPA configurations for Workload that are suitable for horizontal autoscaling
 - [**IdleNode Recommendation**](/docs/tutorials/recommendation/idlenode-recommendation): Find the idle nodes in cluster
+- [**Service Recommendation**](/zh-cn/docs/tutorials/recommendation/service-recommendation): Find the idle load balancer service in cluster
+- [**PV Recommendation**](/zh-cn/docs/tutorials/recommendation/pv-recommendation): Find the idle persist volume in cluster
 
 ### Recommender Framework 
 

diff --git a/site/content/en/docs/Tutorials/Recommendation/service-recommendation.md b/site/content/en/docs/Tutorials/Recommendation/service-recommendation.md
@@ -0,0 +1,97 @@
+---
+title: "闲置节点推荐"
+description: "闲置节点推荐功能介绍"
+weight: 15
+---
+
+Service 推荐通过扫描集群中 Service 的运行状况，帮助用户找到闲置的 Kubernetes Service。
+
+## 动机
+
+通常在 Kubernetes 中我们会使用 Service + Workload 来自动创建和管理负载均衡并将负载均衡挂载到应用上，在日常的运营中难免会出现空闲和低利用率的负载均衡，浪费了大量成本，Service 推荐尝试帮助用户找到这部分 Service 来实现成本优化。
+
+## 推荐示例
+
+```yaml
+apiVersion: analysis.crane.io/v1alpha1
+kind: Recommendation
+metadata:
+  annotations:
+    analysis.crane.io/last-start-time: "2023-06-12 11:52:23"
+    analysis.crane.io/message: Success
+    analysis.crane.io/run-number: "7823"
+  creationTimestamp: "2023-06-12T09:44:23Z"
+  labels:
+    analysis.crane.io/recommendation-rule-name: service-rule
+    analysis.crane.io/recommendation-rule-recommender: Service
+    analysis.crane.io/recommendation-rule-uid: 67807cd9-b4c9-4d63-8493-d330ccace364
+    analysis.crane.io/recommendation-target-kind: Service
+    analysis.crane.io/recommendation-target-name: nginx
+    analysis.crane.io/recommendation-target-namespace: crane-system
+    analysis.crane.io/recommendation-target-version: v1
+  name: service-rule-service-cnwt5
+  namespace: crane-system
+  ownerReferences:
+    - apiVersion: analysis.crane.io/v1alpha1
+      blockOwnerDeletion: false
+      controller: false
+      kind: RecommendationRule
+      name: service-rule
+      uid: 67807cd9-b4c9-4d63-8493-d330ccace364
+spec:
+  adoptionType: StatusAndAnnotation
+  completionStrategy:
+    completionStrategyType: Once
+  targetRef:
+    apiVersion: v1
+    kind: Service
+    name: nginx
+    namespace: crane-system
+  type: Service
+status:
+  action: Delete
+  description: It is a Orphan Service, Pod count is 0
+  lastUpdateTime: "2023-06-12T11:52:23Z"
+```
+
+在该示例中：
+
+- 推荐的 TargetRef 指向了 Service：nginx
+- 推荐类型为 Service 推荐
+- action 是 Delete，这里只是给出建议
+
+## 实现原理
+
+Service 推荐按以下步骤完成一次推荐过程：
+
+1. 扫描集群中所有 LoadBalancer 类型的 Service
+2. 如果 Service 对应的 endpoints 中有 Address 或者 NotReadyAddresses，则不是限制的 Service
+3. 依据 Service 推荐中流量相关 metric 检测 Service 是否小于阈值水位，如果小于水位则判定为闲置节点
+
+## 如何验证推荐结果的准确性
+
+以下是判断节点资源阈值水位的 Prom query，验证时把 node 替换成实际的节点名
+
+```go
+// Container network cumulative count of bytes received
+queryFmtNetReceiveBytes = `sum(rate(container_network_receive_bytes_total{namespace="%s",pod=~"%s",container!=""}[3m]))`
+// Container network cumulative count of bytes transmitted
+queryFmtNetTransferBytes = `sum(rate(container_network_transmit_bytes_total{namespace="%s",pod=~"%s",container!=""}[3m]))`
+```
+
+## 支持的资源类型
+
+只支持 Service 类型，目前只会对 LoadBalancer 类型的 Service 进行分析。
+
+## 参数配置
+
+| 配置项      | 默认值 | 描述                              |
+|----------|-----|---------------------------------|
+| net-receive-bytes | 0   | Service 对应 Pods 接受到的网络请求 bytes，默认不检查 |
+| net-receive-percentile  | 0.99 | 计算接受到的网络请求时的 Percentile         |
+| net-transfer-bytes | 0   | Service 对应 Pods 传输的网络请求 bytes，默认不检查   |
+| net-transfer-percentile | 0.99    | 计算传输的网络请求时的 Percentile          |
+
+注意，当 pod 配置了 liveness/readness probe 后，kubelet 的探测会带来一定的容器流量，因此流量的阈值需要设置的稍微大一些，可结合具体监控数据配置。
+
+如何更新推荐的配置请参考：[**推荐框架**](/zh-cn/docs/tutorials/recommendation/recommendation-framework)
diff --git a/site/content/zh/docs/Tutorials/Recommendation/idlenode-recommendation.md b/site/content/zh/docs/Tutorials/Recommendation/idlenode-recommendation.md
@@ -13,39 +13,46 @@ weight: 15
 ## 推荐示例
 
 ```yaml
-kind: Recommendation
 apiVersion: analysis.crane.io/v1alpha1
+kind: Recommendation
 metadata:
-  name: idlenodes-rule-idlenode-5jxn9
-  namespace: crane-system
+  annotations:
+    analysis.crane.io/last-start-time: "2023-06-09 09:46:33"
+    analysis.crane.io/message: Success
+    analysis.crane.io/run-number: "111"
+  creationTimestamp: "2023-05-31T11:06:10Z"
+  generateName: idlenodes-rule-idlenode-
+  generation: 111
   labels:
     analysis.crane.io/recommendation-rule-name: idlenodes-rule
     analysis.crane.io/recommendation-rule-recommender: IdleNode
-    analysis.crane.io/recommendation-rule-uid: 8921a198-7082-11ed-8b7b-246e960a8d8c
+    analysis.crane.io/recommendation-rule-uid: 25bf5a49-e78f-4f42-8e67-36c0b1b9bb5b
     analysis.crane.io/recommendation-target-kind: Node
     analysis.crane.io/recommendation-target-name: worker-node-1
+    analysis.crane.io/recommendation-target-namespace: ""
     analysis.crane.io/recommendation-target-version: v1
-    beta.kubernetes.io/arch: amd64
-    beta.kubernetes.io/instance-type: bareMetal
-    beta.kubernetes.io/os: linux
+  name: idlenodes-rule-idlenode-px2ck
+  namespace: crane-system
   ownerReferences:
     - apiVersion: analysis.crane.io/v1alpha1
+      blockOwnerDeletion: false
+      controller: false
       kind: RecommendationRule
       name: idlenodes-rule
-      uid: 8921a198-7082-11ed-8b7b-246e960a8d8c
-      controller: false
-      blockOwnerDeletion: false
+      uid: 25bf5a49-e78f-4f42-8e67-36c0b1b9bb5b
 spec:
+  adoptionType: StatusAndAnnotation
+  completionStrategy:
+    completionStrategyType: Once
   targetRef:
+    apiVersion: v1
     kind: Node
     name: worker-node-1
-    apiVersion: v1
   type: IdleNode
-  completionStrategy: {}
 status:
-  targetRef: {}
   action: Delete
-  lastUpdateTime: '2022-11-30T07:46:57Z'
+  description: Node is owned by DaemonSet
+  lastUpdateTime: "2023-06-09T09:46:33Z"
 ```
 
 在该示例中：
@@ -60,4 +67,36 @@ status:
 
 1. 扫描集群中所有节点和节点上的 Pod
 2. 如果节点上所有 Pod 都属于 DaemonSet，则判定为闲置节点
+3. 依据 IdleNode 的其他配置检测节点是否小于阈值水位，如果小于水位则判定为闲置节点
+
+## 如何验证推荐结果的准确性
+
+以下是判断节点资源阈值水位的 Prom query，验证时把 node 替换成实际的节点名
+
+```go
+    // NodeCpuRequestUtilizationExprTemplate is used to query node cpu request utilization by promql, param is node name, node name which prometheus scrape
+NodeCpuRequestUtilizationExprTemplate = `sum(kube_pod_container_resource_requests{node="%s", resource="cpu", unit="core"} * on (node) group_left() max(kube_node_labels{label_beta_kubernetes_io_instance_type!~"eklet", label_node_kubernetes_io_instance_type!~"eklet"}) by (node)) by (node) / sum(kube_node_status_capacity{node="%s", resource="cpu", unit="core"} * on (node) group_left() max(kube_node_labels{label_beta_kubernetes_io_instance_type!~"eklet", label_node_kubernetes_io_instance_type!~"eklet"}) by (node)) by (node) `
+// NodeMemRequestUtilizationExprTemplate is used to query node memory request utilization by promql, param is node name, node name which prometheus scrape
+NodeMemRequestUtilizationExprTemplate = `sum(kube_pod_container_resource_requests{node="%s", resource="memory", unit="byte", namespace!=""} * on (node) group_left() max(kube_node_labels{label_beta_kubernetes_io_instance_type!~"eklet", label_node_kubernetes_io_instance_type!~"eklet"}) by (node)) by (node) / sum(kube_node_status_capacity{node="%s", resource="memory", unit="byte"} * on (node) group_left() max(kube_node_labels{label_beta_kubernetes_io_instance_type!~"eklet", label_node_kubernetes_io_instance_type!~"eklet"}) by (node)) by (node) `
+// NodeCpuUsageUtilizationExprTemplate is used to query node memory usage utilization by promql, param is node name, node name which prometheus scrape
+NodeCpuUsageUtilizationExprTemplate = `sum(label_replace(irate(container_cpu_usage_seconds_total{instance="%s", container!="POD", container!="",image!=""}[1h]), "node", "$1", "instance",  "(^[^:]+)") * on (node) group_left() max(kube_node_labels{label_beta_kubernetes_io_instance_type!~"eklet", label_node_kubernetes_io_instance_type!~"eklet"}) by (node)) by (node) / sum(kube_node_status_capacity{node="%s", resource="cpu", unit="core"} * on (node) group_left() max(kube_node_labels{label_beta_kubernetes_io_instance_type!~"eklet", label_node_kubernetes_io_instance_type!~"eklet"}) by (node)) by (node) `
+// NodeMemUsageUtilizationExprTemplate is used to query node memory usage utilization by promql, param is node name, node name which prometheus scrape
+NodeMemUsageUtilizationExprTemplate = `sum(label_replace(container_memory_usage_bytes{instance="%s", namespace!="",container!="POD", container!="",image!=""}, "node", "$1", "instance", "(^[^:]+)") * on (node) group_left() max(kube_node_labels{label_beta_kubernetes_io_instance_type!~"eklet", label_node_kubernetes_io_instance_type!~"eklet"}) by (node)) by (node) / sum(kube_node_status_capacity{node="%s", resource="memory", unit="byte"} * on (node) group_left() max(kube_node_labels{label_beta_kubernetes_io_instance_type!~"eklet", label_node_kubernetes_io_instance_type!~"eklet"}) by (node)) by (node) `
+```
+
+## 支持的资源类型
+
+只支持 Node，由于 Node 是 Cluster Scope 资源，因此 IdleNode 类型的 Recommendation 均在 crane-system namespace。
+
+## 参数配置
+
+| 配置项      | 默认值  | 描述                                       |
+|----------|------|------------------------------------------|
+| cpu-request-utilization | 0    | 高于该值利用率的节点不是闲置节点，0.5代表50%，默认不检查          |
+| cpu-usage-utilization  | 0    | 高于该值 request 使用率的节点不是闲置节点，0.5代表50%，默认不检查 |
+| cpu-percentile | 0.99 | 计算 cpu 负载时的 Percentile                   |
+| memory-request-utilization | 0    | 高于该值利用率的节点不是闲置节点，0.5代表50%，默认不检查          |
+| memory-usage-utilization  | 0    | 高于该值 request 使用率的节点不是闲置节点，0.5代表50%，默认不检查 |
+| memory-percentile | 0.99 | 计算 memory 负载时的 Percentile                |
 
+如何更新推荐的配置请参考：[**推荐框架**](/zh-cn/docs/tutorials/recommendation/recommendation-framework)
diff --git a/site/content/zh/docs/Tutorials/Recommendation/pv-recommendation.md b/site/content/zh/docs/Tutorials/Recommendation/pv-recommendation.md
@@ -0,0 +1,37 @@
+---
+title: "PV 推荐"
+description: "PV 推荐功能介绍"
+weight: 15
+---
+
+PV 推荐通过扫描集群中 PV 的运行状况，帮助用户找到闲置的 Kubernetes PV。
+
+## 动机
+
+通常在 Kubernetes 中我们会使用 PV + Workload 来自动创建和管理存储卷并将存储卷挂载到应用上，在日常的运营中难免会出现空闲或者空跑的存储卷，浪费了大量成本， PV 推荐尝试帮助用户找到这部分 PV 来实现成本优化。
+
+## 推荐示例
+
+```yaml
+
+```
+
+在该示例中：
+
+- 推荐的 TargetRef 指向了 PV:
+- 推荐类型为 PV 推荐
+- action 是 Delete，这里只是给出建议
+
+## 实现原理
+
+PV 推荐按以下步骤完成一次推荐过程：
+
+1. 扫描集群中所有 PV，找到 PV 对应的 Pod 列表
+2. 如果 PV 没有对应的 PVC，则判断为闲置 PV
+3. 如果没有 Pod 关联这个 PV 和 PVC，则判断为闲置 PVC
+
+## 参数配置
+
+目前 PV 推荐没有参数配置。
+
+如何更新推荐的配置请参考：[**推荐框架**](/zh-cn/docs/tutorials/recommendation/recommendation-framework)
diff --git a/site/content/zh/docs/Tutorials/Recommendation/recommendation-framework.md b/site/content/zh/docs/Tutorials/Recommendation/recommendation-framework.md
@@ -128,6 +128,8 @@ patchData=`kubectl get recommend workloads-rule-replicas-rckvb -n default -o jso
 - [**副本数推荐**](/zh-cn/docs/tutorials/recommendation/replicas-recommendation): 通过 HPA 算法分析应用的真实用量推荐更合适的副本数量
 - [**HPA 推荐**](/zh-cn/docs/tutorials/recommendation/hpa-recommendation): 扫描集群中的 Workload，针对适合适合水平弹性的 Workload 推荐 HPA 配置
 - [**闲置节点推荐**](/zh-cn/docs/tutorials/recommendation/idlenode-recommendation): 扫描集群中的闲置节点
+- [**Service 推荐**](/zh-cn/docs/tutorials/recommendation/service-recommendation): 扫描集群中的闲置 Service
+- [**PV 推荐**](/zh-cn/docs/tutorials/recommendation/pv-recommendation): 扫描集群中的闲置 PV
 
 ### Recommender 框架
 

diff --git a/site/content/zh/docs/Tutorials/Recommendation/service-recommendation.md b/site/content/zh/docs/Tutorials/Recommendation/service-recommendation.md
@@ -0,0 +1,97 @@
+---
+title: "闲置节点推荐"
+description: "闲置节点推荐功能介绍"
+weight: 15
+---
+
+Service 推荐通过扫描集群中 Service 的运行状况，帮助用户找到闲置的 Kubernetes Service。
+
+## 动机
+
+通常在 Kubernetes 中我们会使用 Service + Workload 来自动创建和管理负载均衡并将负载均衡挂载到应用上，在日常的运营中难免会出现空闲和低利用率的负载均衡，浪费了大量成本，Service 推荐尝试帮助用户找到这部分 Service 来实现成本优化。
+
+## 推荐示例
+
+```yaml
+apiVersion: analysis.crane.io/v1alpha1
+kind: Recommendation
+metadata:
+  annotations:
+    analysis.crane.io/last-start-time: "2023-06-12 11:52:23"
+    analysis.crane.io/message: Success
+    analysis.crane.io/run-number: "7823"
+  creationTimestamp: "2023-06-12T09:44:23Z"
+  labels:
+    analysis.crane.io/recommendation-rule-name: service-rule
+    analysis.crane.io/recommendation-rule-recommender: Service
+    analysis.crane.io/recommendation-rule-uid: 67807cd9-b4c9-4d63-8493-d330ccace364
+    analysis.crane.io/recommendation-target-kind: Service
+    analysis.crane.io/recommendation-target-name: nginx
+    analysis.crane.io/recommendation-target-namespace: crane-system
+    analysis.crane.io/recommendation-target-version: v1
+  name: service-rule-service-cnwt5
+  namespace: crane-system
+  ownerReferences:
+    - apiVersion: analysis.crane.io/v1alpha1
+      blockOwnerDeletion: false
+      controller: false
+      kind: RecommendationRule
+      name: service-rule
+      uid: 67807cd9-b4c9-4d63-8493-d330ccace364
+spec:
+  adoptionType: StatusAndAnnotation
+  completionStrategy:
+    completionStrategyType: Once
+  targetRef:
+    apiVersion: v1
+    kind: Service
+    name: nginx
+    namespace: crane-system
+  type: Service
+status:
+  action: Delete
+  description: It is a Orphan Service, Pod count is 0
+  lastUpdateTime: "2023-06-12T11:52:23Z"
+```
+
+在该示例中：
+
+- 推荐的 TargetRef 指向了 Service：nginx
+- 推荐类型为 Service 推荐
+- action 是 Delete，这里只是给出建议
+
+## 实现原理
+
+Service 推荐按以下步骤完成一次推荐过程：
+
+1. 扫描集群中所有 LoadBalancer 类型的 Service
+2. 如果 Service 对应的 endpoints 中有 Address 或者 NotReadyAddresses，则不是限制的 Service
+3. 依据 Service 推荐中流量相关 metric 检测 Service 是否小于阈值水位，如果小于水位则判定为闲置节点
+
+## 如何验证推荐结果的准确性
+
+以下是判断节点资源阈值水位的 Prom query，验证时把 node 替换成实际的节点名
+
+```go
+// Container network cumulative count of bytes received
+queryFmtNetReceiveBytes = `sum(rate(container_network_receive_bytes_total{namespace="%s",pod=~"%s",container!=""}[3m]))`
+// Container network cumulative count of bytes transmitted
+queryFmtNetTransferBytes = `sum(rate(container_network_transmit_bytes_total{namespace="%s",pod=~"%s",container!=""}[3m]))`
+```
+
+## 支持的资源类型
+
+只支持 Service 类型，目前只会对 LoadBalancer 类型的 Service 进行分析。
+
+## 参数配置
+
+| 配置项      | 默认值 | 描述                              |
+|----------|-----|---------------------------------|
+| net-receive-bytes | 0   | Service 对应 Pods 接受到的网络请求 bytes，默认不检查 |
+| net-receive-percentile  | 0.99 | 计算接受到的网络请求时的 Percentile         |
+| net-transfer-bytes | 0   | Service 对应 Pods 传输的网络请求 bytes，默认不检查   |
+| net-transfer-percentile | 0.99    | 计算传输的网络请求时的 Percentile          |
+
+注意，当 pod 配置了 liveness/readness probe 后，kubelet 的探测会带来一定的容器流量，因此流量的阈值需要设置的稍微大一些，可结合具体监控数据配置。
+
+如何更新推荐的配置请参考：[**推荐框架**](/zh-cn/docs/tutorials/recommendation/recommendation-framework)