github.com/gocrane/crane@v0.11.0/docs/installation.zh.md (about) 1 # 产品部署指南 2 3 为了让您更快的部署 Crane ，本文档提供清晰的： 4 5 * 部署环境要求 6 * 具体安装步骤 7 8 Crane 安装时间在10分钟左右，具体时间也依赖集群规模以及硬件能力。目前安装已经非常成熟，如果您安装中遇到任何问题，可以采取如下几种方式： 9 10 * 请首先检查后文的 F&Q 11 * 可以提出一个 [Issue](https://github.com/gocrane/crane/issues/new?assignees=&labels=kind%2Fbug&template=bug_report.md&title=)，我们会认真对待每一个 [Issue](https://github.com/gocrane/crane/issues) 12 13 ## 部署环境要求 14 15 - Kubernetes 1.18+ 16 - Helm 3.1.0 17 18 ## 安装流程 19 20 ### 安装 Helm 21 22 建议参考 Helm 官网[安装文档](https://helm.sh/docs/intro/install/)。 23 24 ### 安装 Prometheus 和 Grafana 25 26 使用 Helm 安装 Prometheus 和 Grafana。 27 28 !!! Note "注意" 29 如果您已经在环境中部署了 Prometheus 和 Grafana，可以跳过该步骤。 30 31 !!! Warning "网络问题" 32 如果你的网络无法访问GitHub资源(GitHub Release, GitHub Raw Content `raw.githubusercontent.com`)。 33 34 那么你可以尝试镜像仓库。但镜像仓库具有一定的**时延**。[镜像仓库](mirror.zh.md) 35 36 37 Crane 使用 Prometheus 抓取集群工作负载对资源的使用情况。安装 Prometheus： 38 39 === "Main" 40 41 ```bash 42 helm repo add prometheus-community https://prometheus-community.github.io/helm-charts 43 helm install prometheus -n crane-system \ 44 --set pushgateway.enabled=false \ 45 --set alertmanager.enabled=false \ 46 --set server.persistentVolume.enabled=false \ 47 -f https://raw.githubusercontent.com/gocrane/helm-charts/main/integration/prometheus/override_values.yaml \ 48 --create-namespace prometheus-community/prometheus 49 ``` 50 51 === "Mirror" 52 53 ```bash 54 helm repo add prometheus-community https://finops-helm.pkg.coding.net/gocrane/prometheus-community 55 helm install prometheus -n crane-system \ 56 --set pushgateway.enabled=false \ 57 --set alertmanager.enabled=false \ 58 --set server.persistentVolume.enabled=false \ 59 -f https://gitee.com/finops/helm-charts/raw/main/integration/prometheus/override_values.yaml \ 60 --create-namespace prometheus-community/prometheus 61 ``` 62 63 64 Crane 的 Fadvisor 使用 Grafana 展示成本预估。安装 Grafana： 65 66 === "Main" 67 68 ```bash 69 helm repo add grafana https://grafana.github.io/helm-charts 70 helm install grafana \ 71 -f https://raw.githubusercontent.com/gocrane/helm-charts/main/integration/grafana/override_values.yaml \ 72 -n crane-system \ 73 --create-namespace grafana/grafana 74 ``` 75 76 === "Mirror" 77 78 ```bash 79 helm repo add grafana https://finops-helm.pkg.coding.net/gocrane/grafana 80 helm install grafana \ 81 -f https://gitee.com/finops/helm-charts/raw/main/integration/grafana/override_values.yaml \ 82 -n crane-system \ 83 --create-namespace grafana/grafana 84 ``` 85 86 ### 安装 Crane 和 Fadvisor 87 88 === "Main" 89 90 ```bash 91 helm repo add crane https://gocrane.github.io/helm-charts 92 helm install crane -n crane-system --create-namespace crane/crane 93 helm install fadvisor -n crane-system --create-namespace crane/fadvisor 94 ``` 95 96 === "Mirror" 97 98 ```bash 99 helm repo add crane https://finops-helm.pkg.coding.net/gocrane/gocrane 100 helm install crane -n crane-system --create-namespace crane/crane 101 helm install fadvisor -n crane-system --create-namespace crane/fadvisor 102 ``` 103 104 ### 安装 Crane-scheduler（可选） 105 ```console 106 helm install scheduler -n crane-system --create-namespace crane/scheduler 107 ``` 108 109 ## 验证安装是否成功 110 111 使用如下命令检查安装的 Deployment 是否正常： 112 113 ```console 114 kubectl get deploy -n crane-system 115 ``` 116 117 结果类似如下： 118 119 ```shell 120 NAME READY UP-TO-DATE AVAILABLE AGE 121 craned 1/1 1 1 31m 122 fadvisor 1/1 1 1 41m 123 grafana 1/1 1 1 42m 124 metric-adapter 1/1 1 1 31m 125 prometheus-kube-state-metrics 1/1 1 1 43m 126 prometheus-server 1/1 1 1 43m 127 ``` 128 129 可以查看本篇[文档](https://github.com/gocrane/helm-charts/blob/main/charts/crane/README.md)获取更多有关 Crane Helm Chart 的信息。 130 131 ## Access Dashboard 132 133 You can use the dashboard to view and manage crane manifests. 134 135  136 137 ### Port Forward 138 139 Easy access to the dashboard through `kubectl port-forward`. 140 141 ```bash 142 kubectl -n crane-system port-forward service/craned 9090:9090 143 ``` 144 145 ### NodePort 146 147 ```bash 148 # Change service type 149 kubectl patch svc craned -n crane-system -p '{"spec": {"type": "NodePort"}}' 150 ``` 151 152 ```bash 153 # Get Dashboard link base on your cluster configuration 154 PORT=$(kubectl get svc -n crane-system craned -o jsonpath='{.spec.ports[?(@.name == "dashboard-service")].nodePort}') 155 NODE_IP=$(kubectl get node -ojsonpath='{.items[].status.addresses[?(@.type == "InternalIP")].address}') 156 echo "Dashboard link: http://${NODE_IP}:${PORT}" 157 ``` 158 159 ### LoadBalancer 160 161 162 #### Quick Start 163 164 ```bash 165 # Change service type 166 kubectl patch svc craned -n crane-system -p '{"spec": {"type": "LoadBalancer"}}' 167 ``` 168 169 #### Example 170 171 ```log 172 $ kubectl patch svc craned -n crane-system -p '{"spec": {"type": "LoadBalancer"}}' 173 174 service/craned patched 175 176 $ kubectl get svc -n crane-system craned 177 NAME TYPE CLUSTER-IP EXTERNAL-IP PORT(S) AGE 178 craned LoadBalancer 10.101.123.74 10.200.0.4 443:30908/TCP,8082:32426/TCP,9090:31331/TCP,8080:31072/TCP 57m 179 180 # Access dashboard via 10.200.0.4:9090 181 ``` 182 183 ### Ingress 184 185 #### kubernetes/ingress-nginx 186 187 If the cluster version is < 1.19, you can create the ingress resources like this: 188 189 ```yaml 190 apiVersion: networking.k8s.io/v1beta1 191 kind: Ingress 192 metadata: 193 name: ingress-crane-dashboard 194 namespace: crane-system 195 spec: 196 ingressClassName: nginx 197 rules: 198 - host: dashboard.gocrane.io # change to your domain 199 http: 200 paths: 201 - path: / 202 backend: 203 serviceName: craned 204 servicePort: 9090 205 ``` 206 207 If the cluster uses Kubernetes version >= 1.19.x, then its suggested to create the second ingress resources, using yaml examples shown below. 208 209 These examples are in conformity with the `networking.kubernetes.io/v1` api. 210 211 ```yaml 212 apiVersion: networking.k8s.io/v1 213 kind: Ingress 214 metadata: 215 name: ingress-crane-dashboard 216 namespace: crane-system 217 spec: 218 rules: 219 - host: dashboard.gocrane.io # change to your domain 220 http: 221 paths: 222 - path: / 223 pathType: Prefix 224 backend: 225 service: 226 name: craned 227 port: 228 number: 9090 229 ingressClassName: nginx 230 ``` 231 232 Example: 233 234 ```log 235 $ kubectl get svc -n ingress-nginx 236 NAME TYPE CLUSTER-IP EXTERNAL-IP PORT(S) AGE 237 ingress-nginx-controller LoadBalancer 10.102.235.229 10.200.0.5 80:32568/TCP,443:30144/TCP 91m 238 ingress-nginx-controller-admission ClusterIP 10.102.49.240 <none> 443/TCP 91m 239 240 $ curl -H "Host: dashboard.gocrane.io" 10.200.0.5 241 <!DOCTYPE html> 242 <html lang="en"> 243 <head> 244 <meta charset="UTF-8" /> 245 <meta name="viewport" content="width=device-width, initial-scale=1.0" /> 246 <title>Crane Dashboard</title> 247 ................................................................ 248 ``` 249 250 #### Traefik 251 252 ```yaml 253 apiVersion: traefik.containo.us/v1alpha1 254 kind: IngressRoute 255 metadata: 256 name: dashboard-crane-ingress 257 namespace: crane-system 258 spec: 259 entryPoints: 260 - web 261 routes: 262 - kind: Rule 263 match: Host(`dashboard.gocrane.io`) 264 services: 265 - name: craned 266 port: 9090 267 ``` 268 269 ```log 270 $ kubectl get svc -n traefik-v2 271 NAME TYPE CLUSTER-IP EXTERNAL-IP PORT(S) AGE 272 traefik LoadBalancer 10.107.109.44 10.200.0.6 80:30102/TCP,443:30139/TCP 16m 273 274 $ curl -H "Host: dashboard.gocrane.io" 10.200.0.6 275 <!DOCTYPE html> 276 <html lang="en"> 277 <head> 278 <meta charset="UTF-8" /> 279 <meta name="viewport" content="width=device-width, initial-scale=1.0" /> 280 <title>Crane Dashboard</title> 281 ................................................................ 282 ``` 283 284 ## 自定义安装 285 286 通过 YAML 安装 `Crane` 。 287 288 === "Main" 289 290 ```bash 291 git clone https://github.com/gocrane/crane.git 292 CRANE_LATEST_VERSION=$(curl -s https://api.github.com/repos/gocrane/crane/releases/latest | grep -oP '"tag_name": "\K(.*)(?=")') 293 git checkout $CRANE_LATEST_VERSION 294 kubectl apply -f deploy/manifests 295 kubectl apply -f deploy/craned 296 kubectl apply -f deploy/metric-adapter 297 ``` 298 299 === "Mirror" 300 301 ```bash 302 git clone https://e.coding.net/finops/gocrane/crane.git 303 CRANE_LATEST_VERSION=$(curl -s https://api.github.com/repos/gocrane/crane/releases/latest | grep -oP '"tag_name": "\K(.*)(?=")') 304 git checkout $CRANE_LATEST_VERSION 305 kubectl apply -f deploy/manifests 306 kubectl apply -f deploy/craned 307 kubectl apply -f deploy/metric-adapter 308 ``` 309 310 如果您想自定义 Crane 里配置 Prometheus 的 HTTP 地址，请参考以下的命令。如果您在集群里已存在一个 Prometheus，请将 Server 地址填于`CUSTOMIZE_PROMETHEUS` 。 311 312 ```console 313 export CUSTOMIZE_PROMETHEUS= 314 if [ $CUSTOMIZE_PROMETHEUS ]; then sed -i '' "s/http:\/\/prometheus-server.crane-system.svc.cluster.local:8080/${CUSTOMIZE_PROMETHEUS}/" deploy/craned/deployment.yaml ; fi 315 ``` 316 317 ## 安装常见问题 318 319 ### 安装 Crane 报错 320 321 当您执行 `helm install crane -n crane-system --create-namespace crane/crane` 命令时，可能会遇到如下错误： 322 323 ```shell 324 Error: rendered manifests contain a resource that already exists. Unable to continue with install: APIService "v1beta1.custom.metrics.k8s.io" in namespace "" exists and cannot be imported into the current release: invalid ownership metadata; label validation error: missing key "app.kubernetes.io/managed-by": must be set to "Helm"; annotation validation error: missing key "meta.helm.sh/release-name": must be set to "crane"; annotation validation error: missing key "meta.helm.sh/release-namespace": must be set to "crane-system" 325 ``` 326 327 原因：集群安装过 custom metric 的 APIService，所以报错。可以把之前的删除再重新执行安装 Crane 的命令，删除方式：`kubectl delete apiservice v1beta1.custom.metrics.k8s.io`。 328 329 ### 获取 Crane URL 的其它方式 330 331 #### NodePort 方式 332 333 您可以将 Crane 的 Service 的类型换成 NodePort 类型，这样可以直接通过集群任意节点 IP + 该服务里dashboard- service 端口号的方式，打开控制台。 334 335 具体操作：修改 crane-system 命名空间下名为 craned 的 Service，将其访问方式该为 NodePort 的方式，然后获取某一集群的节点 IP，以及相应的端口号，端口号如下所示： 336 337  338 339 注意：若您的集群节点只有内网 IP，则访问该 IP 的计算机需要在同一内网。若集群节点拥有外网 IP，则没有相关问题。 340 341 #### LoadBalance 方式 342 343 若您使用的是公有云厂商的服务，您可以将 Crane 的 Service 的类型换成公网 LB 类型，这样可以直接通过 LB IP + 9090 端口号的方式，打开控制台。 344 345 具体操作：修改 crane-system 命名空间下名为 craned 的 Service，将其访问方式该为公网 LB 的方式。