实现prometheus+grafana的监控部署

直接贴部署用的文件信息了

kubectl label node xxx monitoring=true

创建命名空间

kubectl create ns monitoring

部署operator

kubectl apply -f operator-rbac.yml
kubectl apply -f operator-dp.yml
kubectl apply -f operator-crd.yml

定义node-export

kubectl apply -f ./node-exporter/node-exporter-sa.yml
kubectl apply -f ./node-exporter/node-exporter-rbac.yml
kubectl apply -f ./node-exporter/node-exporter-svc.yml
kubectl apply -f ./node-exporter/node-exporter-ds.yml

自定义配置文件,定义显示方式

kubectl apply -f ./grafana/pv-pvc-hostpath.yml
kubectl apply -f ./grafana/grafana-sa.yml
kubectl apply -f ./grafana/grafana-source.yml
kubectl apply -f ./grafana/grafana-datasources.yml
kubectl apply -f ./grafana/grafana-admin-secret.yml
kubectl apply -f ./grafana/grafana-svc.yml

创建配置conifgmap

kubectl create configmap grafana-config --from-file=./grafana/grafana.ini --namespace=monitoring
kubectl create configmap all-grafana-dashboards --from-file=./grafana/dashboard --namespace=monitoring


kubectl apply -f ./grafana/grafana-dp.yml
kubectl apply -f ./service-discovery/kube-controller-manager-svc.yml
kubectl apply -f ./service-discovery/kube-scheduler-svc.yml

自定义配置文件,定义收集和报警规则

kubectl apply -f ./prometheus/prometheus-secret.yml
kubectl apply -f ./prometheus/prometheus-rules.yml
kubectl apply -f ./prometheus/prometheus-rbac.yml
kubectl apply -f ./prometheus/prometheus-svc.yml

prometheus-operator 部署成功后才能创建成功

kubectl apply -f ./prometheus/pv-pvc-hostpath.yaml
kubectl apply -f ./prometheus/prometheus-main.yml

监控目标,lable必须是k8s-app 因为prometheus是按这个查找的。不然prometheus采集不了该

kubectl apply -f ./servicemonitor/alertmanager-sm.yml
kubectl apply -f ./servicemonitor/coredns-sm.yml
kubectl apply -f ./servicemonitor/kube-apiserver-sm.yml
kubectl apply -f ./servicemonitor/kube-controller-manager-sm.yml
kubectl apply -f ./servicemonitor/kube-scheduler-sm.yml
kubectl apply -f ./servicemonitor/kubelet-sm.yml
kubectl apply -f ./servicemonitor/kubestate-metrics-sm.yml
kubectl apply -f ./servicemonitor/node-exporter-sm.yml
kubectl apply -f ./servicemonitor/prometheus-operator-sm.yml
kubectl apply -f ./servicemonitor/prometheus-sm.yml
kubectl apply -f ./servicemonitor/pushgateway-sm.yml

prometheus-adapter 部署

kubectl apply -f ./prometheus_adapter/metric_rule.yaml
kubectl apply -f ./prometheus_adapter/prometheus_adapter.yaml

受限于篇幅就不张贴部署脚本内容,详情请参见

GitHub - chenrui2200/prometheus_k8s_install

创建出进监控pod

node_exporterprometheus_k8s 是在 Kubernetes 环境中监控和收集指标的两个重要组件。它们之间有着密切的关系,以下是它们的详细讲解及相互关系。

Node Exporter

  • 定义 : node_exporter 是 Prometheus 官方提供的一个工具,用于收集和暴露操作系统及硬件的性能指标。这些指标包括 CPU、内存、磁盘、网络等系统级别的性能数据。

  • 功能:

    • 它运行在每个节点上,监控该节点的系统资源使用情况。
    • 提供的指标格式符合 Prometheus 的要求,因此可以直接被 Prometheus 抓取。
  • 使用场景:

    • 适用于监控物理机、虚拟机或 Kubernetes 节点的基础设施健康状态。

Prometheus K8s

  • 定义 : prometheus_k8s 是一个 Prometheus 实例,专门用于在 Kubernetes 集群中监控 Kubernetes 资源及其运行的应用。

  • 功能:

    • 通过 Kubernetes 的 API,自动发现集群中的服务和容器,并抓取它们暴露的指标。
    • 可以监控 Kubernetes 组件(如 kube-apiserver、kube-scheduler、kube-controller-manager)及各个应用程序的性能。
  • 使用场景:

    • 适用于监控整个 Kubernetes 集群的健康状况和性能。

二者的关系

  1. 数据来源:

    • node_exporter 收集每个节点的系统性能指标,然后 Prometheus 可以抓取这些指标。
    • prometheus_k8s 则从 Kubernetes 中的其他组件和应用程序收集指标。
  2. 监控层次:

    • node_exporter 主要关注底层硬件和操作系统级别的监控。
    • prometheus_k8s 关注 Kubernetes 资源和应用的监控,包括 Pods、服务和其它 Kubernetes 对象。
  3. 集成:

    • 在 Kubernetes 环境中,通常会在每个节点上运行 node_exporter,并配置 prometheus_k8s 来定期抓取 node_exporter 的指标,这样就能实现对节点性能的全面监控。

node_exporterprometheus_k8s 是相辅相成的。node_exporter 提供了基础设施级别的监控,而 prometheus_k8s 则补充了 Kubernetes 资源和应用的监控。结合使用这两个组件,能够实现对整个系统的全面监控,帮助运维人员及时发现和解决问题。

查看前端展示

python实现prometheus客户端

import json, datetime, time
import requests
import pysnooper


class Prometheus():

    def __init__(self, host=''):
        #  '/api/v1/query_range'    查看范围数据
        #  '/api/v1/query'    瞬时数据查询
        self.host = host
        self.query_path = 'http://%s/api/v1/query' % self.host
        self.query_range_path = 'http://%s/api/v1/query_range' % self.host

    # @pysnooper.snoop()
    def get_istio_service_metric(self, namespace):
        service_metric = {
            "qps": {},
            "gpu": {},
            "memory": {},
            "cpu": {}
        }
        # qps请求
        mem_expr = 'sum by (destination_workload,response_code) (irate(istio_requests_total{destination_service_namespace="%s"}[1m]))' % (namespace,)
        # print(mem_expr)
        params = {
            'query': mem_expr,
            'start': int(time.time())-300,
            'end': int(time.time()),
            'step': "1m",  # 运行小于1分钟的,将不会被采集到
            # 'timeout':"30s"
        }
        print(params)

        try:
            res = requests.get(url=self.query_range_path, params=params)
            metrics = json.loads(res.content.decode('utf8', 'ignore'))
            if metrics['status'] == 'success':
                metrics = metrics['data']['result']
                if metrics:
                    for service in metrics:
                        service_name = service['metric']['destination_workload']
                        if service_name not in service_metric['qps']:
                            service_metric['qps'][service_name] = {}
                        service_metric["qps"][service_name] = service['values']

        except Exception as e:
            print(e)

        # 内存
        mem_expr = 'sum by (pod) (container_memory_working_set_bytes{job="kubelet", image!="",container_name!="POD",namespace="%s"})' % (namespace,)
        # print(mem_expr)
        params = {
            'query': mem_expr,
            'start': int(time.time()) - 300,
            'end': int(time.time()),
            'step': "1m",  # 运行小于1分钟的,将不会被采集到
            # 'timeout':"30s"
        }
        print(params)

        try:
            res = requests.get(url=self.query_range_path, params=params)
            metrics = json.loads(res.content.decode('utf8', 'ignore'))
            if metrics['status'] == 'success':
                metrics = metrics['data']['result']
                if metrics:
                    for pod in metrics:
                        pod_name = pod['metric']['pod']
                        if pod_name not in service_metric['memory']:
                            service_metric[pod_name] = {}
                        service_metric['memory'][pod_name] = pod['values']

        except Exception as e:
            print(e)

        # cpu获取
        cpu_expr = "sum by (pod) (rate(container_cpu_usage_seconds_total{namespace='%s',container!='POD'}[1m]))" % (namespace)

        params = {
            'query': cpu_expr,
            'start': int(time.time()) - 300,
            'end': int(time.time()),
            'step': "1m",  # 运行小于1分钟的,将不会被采集到
            # 'timeout':"30s"
        }
        print(params)
        try:

            res = requests.get(url=self.query_range_path, params=params)
            metrics = json.loads(res.content.decode('utf8', 'ignore'))
            if metrics['status'] == 'success':
                metrics = metrics['data']['result']
                if metrics:
                    for pod in metrics:
                        pod_name = pod['metric']['pod']
                        if pod_name not in service_metric['cpu']:
                            service_metric[pod_name] = {}
                        service_metric['cpu'][pod_name] = pod['values']

        except Exception as e:
            print(e)

        gpu_expr = "avg by (pod) (DCGM_FI_DEV_GPU_UTIL{namespace='%s'})" % (namespace)

        params = {
            'query': gpu_expr,
            'start': (datetime.datetime.now() - datetime.timedelta(days=1) - datetime.timedelta(hours=8)).strftime('%Y-%m-%dT%H:%M:%S.000Z'),
            'end': datetime.datetime.now().strftime('%Y-%m-%dT%H:%M:%S.000Z'),
            'step': "1m",  # 运行小于1分钟的,将不会被采集到
            # 'timeout':"30s"
        }
        print(params)
        try:

            res = requests.get(url=self.query_range_path, params=params)
            metrics = json.loads(res.content.decode('utf8', 'ignore'))
            if metrics['status'] == 'success':
                metrics = metrics['data']['result']
                if metrics:
                    # print(metrics)
                    for pod in metrics:
                        pod_name = pod['metric']['pod']
                        if pod_name not in service_metric['gpu']:
                            service_metric['gpu'][pod_name] = {}
                        service_metric['gpu'][pod_name] = pod['values']

        except Exception as e:
            print(e)

        return service_metric



    # 获取当前pod利用率
    # @pysnooper.snoop()
    def get_resource_metric(self):
        max_cpu = 0
        max_mem = 0
        ave_gpu = 0
        pod_metric = {}
        # 这个pod  30分钟内的最大值
        mem_expr = "sum by (pod) (container_memory_working_set_bytes{container!='POD', container!=''})"
        # print(mem_expr)
        params = {
            'query': mem_expr,
            'start': int(time.time()) - 300,
            'end': int(time.time()),
            'step': "1m",  # 运行小于1分钟的,将不会被采集到
            # 'timeout':"30s"
        }
        print(params)

        try:
            res = requests.get(url=self.query_range_path, params=params)
            metrics = json.loads(res.content.decode('utf8', 'ignore'))
            if metrics['status'] == 'success':
                metrics = metrics['data']['result']
                if metrics:
                    for pod in metrics:
                        if pod['metric']:
                            pod_name = pod['metric']['pod']
                            values = max([float(x[1]) for x in pod['values']])
                            if pod_name not in pod_metric:
                                pod_metric[pod_name] = {}
                            pod_metric[pod_name]['memory'] = round(values / 1024 / 1024 / 1024, 2)

        except Exception as e:
            print(e)

        cpu_expr = "sum by (pod) (rate(container_cpu_usage_seconds_total{container!='POD'}[1m]))"

        params = {
            'query': cpu_expr,
            'start': int(time.time()) - 300,
            'end': int(time.time()),
            'step': "1m",  # 运行小于1分钟的,将不会被采集到
            # 'timeout':"30s"
        }
        print(params)
        try:

            res = requests.get(url=self.query_range_path, params=params)
            metrics = json.loads(res.content.decode('utf8', 'ignore'))
            if metrics['status'] == 'success':
                metrics = metrics['data']['result']
                if metrics:
                    for pod in metrics:
                        if pod['metric']:
                            pod_name = pod['metric']['pod']
                            values = [float(x[1]) for x in pod['values']]
                            # values = round(sum(values) / len(values), 2)
                            values = round(max(values), 2)
                            if pod_name not in pod_metric:
                                pod_metric[pod_name] = {}
                            pod_metric[pod_name]['cpu'] = values

        except Exception as e:
            print(e)

        gpu_expr = "avg by (pod) (DCGM_FI_DEV_GPU_UTIL)"

        params = {
            'query': gpu_expr,
            'start': int(time.time()) - 300,
            'end': int(time.time()),
            'step': "1m",  # 运行小于1分钟的,将不会被采集到
            # 'timeout':"30s"
        }
        print(params)
        try:

            res = requests.get(url=self.query_range_path, params=params)
            metrics = json.loads(res.content.decode('utf8', 'ignore'))
            if metrics['status'] == 'success':
                metrics = metrics['data']['result']
                if metrics:
                    # print(metrics)
                    for pod in metrics:
                        if pod['metric']:
                            pod_name = pod['metric']['pod']
                            values = [float(x[1]) for x in pod['values']]
                            # values = round(sum(values)/len(values),2)
                            values = round(max(values), 2)
                            if pod_name not in pod_metric:
                                pod_metric[pod_name] = {}
                            pod_metric[pod_name]['gpu'] = values / 100

        except Exception as e:
            print(e)

        return pod_metric

    # @pysnooper.snoop()
    def get_namespace_resource_metric(self, namespace):
        max_cpu = 0
        max_mem = 0
        ave_gpu = 0
        pod_metric = {}
        # 这个pod  30分钟内的最大值
        mem_expr = "sum by (pod) (container_memory_working_set_bytes{namespace='%s',container!='POD', container!=''})" % (namespace,)
        # print(mem_expr)
        params = {
            'query': mem_expr,
            'start': int(time.time()) - 60*60*24,
            'end': int(time.time()),
            'step': "1m",  # 运行小于1分钟的,将不会被采集到
            # 'timeout':"30s"
        }
        print(params)

        try:
            res = requests.get(url=self.query_range_path, params=params)
            metrics = json.loads(res.content.decode('utf8', 'ignore'))
            if metrics['status'] == 'success':
                metrics = metrics['data']['result']
                if metrics:
                    for pod in metrics:
                        pod_name = pod['metric']['pod']
                        values = max([float(x[1]) for x in pod['values']])
                        if pod_name not in pod_metric:
                            pod_metric[pod_name] = {}
                        pod_metric[pod_name]['memory'] = round(values / 1024 / 1024 / 1024, 2)

        except Exception as e:
            print(e)

        cpu_expr = "sum by (pod) (rate(container_cpu_usage_seconds_total{namespace='%s',container!='POD'}[1m]))" % (namespace)

        params = {
            'query': cpu_expr,
            'start': int(time.time()) - 60*60*24,
            'end': int(time.time()),
            'step': "1m",  # 运行小于1分钟的,将不会被采集到
            # 'timeout':"30s"
        }
        print(params)
        try:

            res = requests.get(url=self.query_range_path, params=params)
            metrics = json.loads(res.content.decode('utf8', 'ignore'))
            if metrics['status'] == 'success':
                metrics = metrics['data']['result']
                if metrics:
                    for pod in metrics:
                        pod_name = pod['metric']['pod']
                        values = [float(x[1]) for x in pod['values']]
                        # values = round(sum(values) / len(values), 2)
                        values = round(max(values), 2)
                        if pod_name not in pod_metric:
                            pod_metric[pod_name] = {}
                        pod_metric[pod_name]['cpu'] = values

        except Exception as e:
            print(e)

        gpu_expr = "avg by (pod) (DCGM_FI_DEV_GPU_UTIL{namespace='%s'})" % (namespace)

        params = {
            'query': gpu_expr,
            'start': int(time.time()) - 60*60*24,
            'end': int(time.time()),
            'step': "1m",  # 运行小于1分钟的,将不会被采集到
            # 'timeout':"30s"
        }
        print(params)
        try:

            res = requests.get(url=self.query_range_path, params=params)
            metrics = json.loads(res.content.decode('utf8', 'ignore'))
            if metrics['status'] == 'success':
                metrics = metrics['data']['result']
                if metrics:
                    # print(metrics)
                    for pod in metrics:
                        pod_name = pod['metric']['pod']
                        values = [float(x[1]) for x in pod['values']]
                        # values = round(sum(values)/len(values),2)
                        values = round(max(values), 2)
                        if pod_name not in pod_metric:
                            pod_metric[pod_name] = {}
                        pod_metric[pod_name]['gpu'] = values / 100

        except Exception as e:
            print(e)

        return pod_metric

        # @pysnooper.snoop()

    def get_pod_resource_metric(self, pod_name, namespace):
        max_cpu = 0
        max_mem = 0
        ave_gpu = 0

        # 这个pod  30分钟内的最大值
        mem_expr = "sum by (pod) (container_memory_working_set_bytes{namespace='%s', pod=~'%s.*',container!='POD', container!=''})"%(namespace,pod_name)
        # print(mem_expr)
        params = {
            'query': mem_expr,
            'start': int(time.time()) - 60*60*24,
            'end': int(time.time()),
            'step': "1m",  # 运行小于1分钟的,将不会被采集到
            # 'timeout':"30s"
        }
        print(params)

        try:
            res = requests.get(url=self.query_range_path, params=params)
            metrics = json.loads(res.content.decode('utf8', 'ignore'))
            if metrics['status'] == 'success':
                metrics = metrics['data']['result']
                if metrics:
                    metrics = metrics[0]['values']
                    for metric in metrics:
                        if int(metric[1]) > max_mem:
                            max_mem = int(metric[1]) / 1024 / 1024 / 1024

        except Exception as e:
            print(e)

        cpu_expr = "sum by (pod) (rate(container_cpu_usage_seconds_total{namespace='%s',pod=~'%s.*',container!='POD'}[1m]))" % (namespace, pod_name)

        params = {
            'query': cpu_expr,
            'start': int(time.time()) - 60*60*24,
            'end': int(time.time()),
            'step': "1m",  # 运行小于1分钟的,将不会被采集到
            # 'timeout':"30s"
        }
        print(params)
        try:

            res = requests.get(url=self.query_range_path, params=params)
            metrics = json.loads(res.content.decode('utf8', 'ignore'))
            if metrics['status'] == 'success':
                metrics = metrics['data']['result']
                if metrics:
                    metrics = metrics[0]['values']
                    for metric in metrics:
                        if float(metric[1]) > max_cpu:
                            max_cpu = float(metric[1])
        except Exception as e:
            print(e)

        gpu_expr = "avg by (pod) (DCGM_FI_DEV_GPU_UTIL{namespace='%s',pod=~'%s.*'})" % (namespace, pod_name)

        params = {
            'query': gpu_expr,
            'start': int(time.time()) - 60*60*24,
            'end': int(time.time()),
            'step': "1m",  # 运行小于1分钟的,将不会被采集到
            # 'timeout':"30s"
        }
        print(params)
        try:
            res = requests.get(url=self.query_range_path, params=params)
            metrics = json.loads(res.content.decode('utf8', 'ignore'))
            if metrics['status'] == 'success':
                metrics = metrics['data']['result']
                if metrics:
                    metrics = metrics[0]['values']
                    all_util = [float(metric[1]) for metric in metrics]
                    ave_gpu = sum(all_util) / len(all_util) / 100
        except Exception as e:
            print(e)

        return {"cpu": round(max_cpu, 2), "memory": round(max_mem, 2), 'gpu': round(ave_gpu, 2)}

    # todo 获取机器的负载补充完整
    # @pysnooper.snoop()
    def get_machine_metric(self):
        # 这个pod  30分钟内的最大值
        metrics = {
            "pod_num": "sum(kubelet_running_pod_count)by (node)",
            "request_memory": "",
            "request_cpu": "",
            "request_gpu": "",
            "used_memory": "",
            "used_cpu": "",
            "used_gpu": "",
        }
        back = {}
        for metric_name in metrics:
            # print(mem_expr)
            params = {
                'query': metrics[metric_name],
                'timeout': "30s"
            }
            print(params)
            back[metric_name] = {}

            try:
                res = requests.get(url=self.query_path, params=params)
                metrics = json.loads(res.content.decode('utf8', 'ignore'))
                if metrics['status'] == 'success':
                    metrics = metrics['data']['result']
                    if metrics:
                        for metric in metrics:
                            node = metric['metric']['node']
                            if ':' in node:
                                node = node[:node.index(':')]
                            value = metric['value'][1]
                            back[metric_name][node] = int(value)


            except Exception as e:
                print(e)

        return back

Prometheus 查询语法使用一种类似于函数式编程的表达方式,允许用户从时间序列数据库中提取和聚合数据。你提供的查询语句 cpu_expr 是一个典型的 Prometheus 查询,下面是对其各个部分的详细讲解:

  1. 函数 sum:是一个聚合函数,用于对一组时间序列进行求和。它可以用于计算所有匹配指标的总和。

  2. by (pod) 这是一个分组操作,表示将结果按 pod 标签进行分组。这样,查询结果将显示每个 pod 的 CPU 使用量总和。

  3. rate(container_cpu_usage_seconds_total{...}[1m]):

    rate 是一个用于计算时间序列速率的函数。在这里,它计算 container_cpu_usage_seconds_total 指标在过去 1 分钟内的变化率。container_cpu_usage_seconds_total 是一个计数器类型的指标,表示容器 CPU 使用时间的总和。{namespace='%s',container!='POD'} 是一个标签选择器,用于过滤指标。它选择特定命名空间下的所有容器,但排除了名为 POD 的容器。

  • 整体而言,这个查询的目的是计算指定命名空间下,所有非 POD 容器的 CPU 使用率,并按每个 pod 进行分组。通过使用 rate 函数,它可以提供 CPU 使用的实时速率,而不是总使用时间,这样更能反映容器的当前状态。

      def echart(self, filters=None):
          
          prometheus = Prometheus(conf.get('PROMETHEUS', 'prometheus-k8s.monitoring:9090'))
          # 获取 prometheus-k8s 地址
          pod_resource_metric = prometheus.get_resource_metric()
          print('pod_resource_metric', pod_resource_metric)
    
          all_resource = {
              "mem_all": sum([int(global_cluster_load[cluster_name]['mem_all']) for cluster_name in global_cluster_load]),
              "cpu_all": sum([int(global_cluster_load[cluster_name]['cpu_all']) for cluster_name in global_cluster_load]),
              "gpu_all": sum([int(global_cluster_load[cluster_name]['gpu_all']) for cluster_name in global_cluster_load]),
          }
          all_resource_req = {
              "mem_req": sum([int(global_cluster_load[cluster_name]['mem_req']) for cluster_name in global_cluster_load]),
              "cpu_req": sum([int(global_cluster_load[cluster_name]['cpu_req']) for cluster_name in global_cluster_load]),
              "gpu_req": sum([int(global_cluster_load[cluster_name]['gpu_req']) for cluster_name in global_cluster_load]),
          }
          all_resource_used = {
              "mem_used": sum([pod_resource_metric[x].get('memory', 0) for x in pod_resource_metric]),
              "cpu_used": sum([pod_resource_metric[x].get('cpu', 0) for x in pod_resource_metric]),
              "gpu_used": sum([pod_resource_metric[x].get('gpu', 0) for x in pod_resource_metric]),
          }
    
          option = {
              "title": [
                  {
                      "subtext": __('集群信息'),
                      "MEM_NAME": __('内存请求占有率'),
                      "MEM_VALUE": int(100 * all_resource_req['mem_req'] / (0.001 + all_resource['mem_all'])),
                      "CPU_NAME": __('CPU占用率'),
                      "CPU_VALUE": int(all_resource['cpu_all']),
                      'GPU_NAME': __('GPU总量(卡)'),
                      'GPU_VALUE': int(all_resource['gpu_all']),
                      'MEM_MAX': int(all_resource['mem_all'] * 2),
                      'CPU_MAX': int(all_resource['cpu_all'] * 2),
                      'GPU_MAX': int(all_resource['gpu_all'] * 2)
                  },
                  {
                      "subtext": __('资源占用率'),
                      "MEM_NAME": __('内存占用率'),
                      "MEM_VALUE": int(100 * all_resource_req['mem_req'] / (0.001 + all_resource['mem_all'])),
                      "CPU_NAME": __('CPU占用率'),
                      "CPU_VALUE": int(100 * all_resource_req['cpu_req'] / (0.001 + all_resource['cpu_all'])),
                      "GPU_NAME": __('GPU占用率'),
                      "GPU_VALUE": int(100 * all_resource_req['gpu_req'] / (0.001 + all_resource['gpu_all']))
    
                  },
                  {
                      "subtext": __('资源利用率'),
                      'MEM_NAME': __('内存利用率'),
                      'MEM_VALUE': str(min(100,
                                           int(100 *
                                               all_resource_used[
                                                   'mem_used'] / (
                                                       0.001 +
                                                       all_resource[
                                                           'mem_all'])))),
                      'CPU_NAME': __('CPU利用率'),
                      'CPU_VALUE': str(min(100, int(100 *
                                                    all_resource_used[
                                                        'cpu_used'] / (
                                                            0.001 +
                                                            all_resource[
                                                                'cpu_all'])))),
                      'GPU_NAME': __('GPU利用率'),
                      'GPU_VALUE': str(min(100, int(100 *
                                                    all_resource_used[
                                                        'gpu_used'] / (
                                                            0.001 +
                                                            all_resource[
                                                                'gpu_all']))))
                  }
              ]
          }
    
          return option
    

附一个用上面代码实现的运维表盘界面

相关推荐
陈小肚30 分钟前
k8s 1.28.2 集群部署 Thanos 对接 MinIO 实现 Prometheus 数据长期存储
kubernetes·prometheus·thanos
福大大架构师每日一题4 小时前
27.9 调用go-ansible执行playbook拷贝json文件重载采集器
golang·json·ansible·prometheus
迷茫运维路21 小时前
Prometheus+Telegraf实现自定义监控项配置
运维·prometheus·telegraf·自定义监控
Mr. G K2 天前
【grafana+Prometheus(普罗米修斯)实现监控功能】
grafana·prometheus
再ZzZ2 天前
Docker快速安装Grafana
服务器·docker·grafana·1024程序员节
编程泓哥2 天前
grafana 8.0 添加钉钉告警
钉钉·grafana
Cat_Xu3 天前
【TEST】负载/性能测试工具 Grafana K6 (Docker 版)
测试工具·docker·grafana·数据可视化
迷茫运维路3 天前
golang开发alertmanagerWebhook,实现prometheus+alertmanagerWebhook告警
golang·prometheus·webhook·alertmanager·监控告警
闫小甲8 天前
使用Spring Boot和Micrometer实现交易度量监控
spring boot·prometheus·micrometer
迷茫运维路8 天前
Prometheus运维监控平台之服务发现配置、标签及监控规则编写(二)
运维·服务发现·prometheus