为我的自定义指标收集器公开特定指标位置

Question

我正在编写自定义 k8s 指标收集器，目的是监控两个不同集群中的应用程序版本。如何暴露“/metrics”等指标的具体位置，并在无限循环中收集指标？

那是我的自定义指标收集器示例：

import time
from prometheus_client import start_http_server
from prometheus_client.core import REGISTRY, CounterMetricFamily
from kubernetes import client, config, watch

class CustomCollector(object):
    def __init__(self):
        pass
    def collect(self):
        g = CounterMetricFamily("retail_pods_info", 'info about pods', labels=['secret','namespace','deployment_name','image','helm'])
        config.load_kube_config('config')
        v1 = client.CoreV1Api()
        group = "argoproj.io"
        version = "v1alpha1"
        plural = "applications"
        #kind = "Application"
        namespace = "argo-cd"
        pod_list: client.V1PodList = v1.list_pod_for_all_namespaces(watch=False)
        pods: list[client.V1Pod] = pod_list.items
        metrics_list = []
        for pod in pods:
            metadata: client.V1ObjectMeta = pod.metadata
            spec: client.V1PodSpec = pod.spec
            volumes: list[client.V1Volume] = spec.volumes
            if volumes is not None:
                for volume in volumes:
                    if volume.projected:
                        projected: client.V1ProjectedVolumeSource = volume.projected
                        sources: list[client.V1VolumeProjection] = projected.sources
                        for source in sources:
                            if source.secret:
                                secret: client.V1SecretProjection = source.secret
                                s = secret.name + " " + metadata.namespace.lower() + " " + metadata.name.lower().rsplit('-',2)[0] + " " + pod.spec.containers[0].image
                                metrics_list.append(s.split())
        api_client = client.ApiClient()
        argocd_api = client.CustomObjectsApi(api_client)
        argocd_apps = argocd_api.list_namespaced_custom_object(group, version, namespace, plural, watch=False)
        for metric in metrics_list:
            for app in argocd_apps["items"]:
                if metric[2] == app["metadata"]["name"]:
                    helm_version=app["spec"]["source"]["repoURL"]+"-"+app["spec"]["source"]["targetRevision"]
                    metric.append(helm_version)
                    g.add_metric([metric[0], metric[1], metric[2], metric[3], metric[4]], 1)
                    yield g
        #
        # for k in metrics_list:
        #     g.add_metric([k[0],k[1],k[2],k[3]], 1)
        #     yield g

if __name__ == '__main__':
    start_http_server(8000)
    REGISTRY.register(CustomCollector())
    while True:
        time.sleep(60)

Answer 1

关于端点

/metrics

：
方法

start_http_server

在指定端口上启动一个服务器，该服务器响应任何带有指标的查询。因此，对路径

/metrics

的请求将默认使用生成的度量来响应。

关于“在无限循环中收集指标”：
您的应用程序已经在执行此操作（有点）。由于您注册了自定义收集器，因此每个请求都将调用

collect

方法。而且由于普罗米修斯在收集指标时基本上是在无限循环中进行的 - 你的收集器也会这样做。