Prometheus 操作员警告“资源版本太旧”

Question

我想了解导致错误的资源是什么？这有助于理解是什么触发了这个

普罗米修斯： quay.io/coreos/prometheus-operator：0.38.1 quay.io/prometheus/prometheus: 2.18.2

level=warn ts=2020-09-01T09:23:33.759Z caller=klog.go:86 component=k8s_client_runtime func=Warningf msg="/app/discovery/kubernetes/kubernetes.go:361: watch of *v1.Endpoints ended with: too old resource version: 77932793 (77933474)"
level=warn ts=2020-09-01T09:24:34.683Z caller=klog.go:86 component=k8s_client_runtime func=Warningf msg="/app/discovery/kubernetes/kubernetes.go:361: watch of *v1.Endpoints ended with: too old resource version: 77931607 (77933976)"
level=warn ts=2020-09-01T09:25:59.216Z caller=klog.go:86 component=k8s_client_runtime func=Warningf msg="/app/discovery/kubernetes/kubernetes.go:361: watch of *v1.Endpoints ended with: too old resource version: 77932221 (77934666)"
level=warn ts=2020-09-01T09:27:51.400Z caller=klog.go:86 component=k8s_client_runtime func=Warningf msg="/app/discovery/kubernetes/kubernetes.go:361: watch of *v1.Endpoints ended with: too old resource version: 77932973 (77935587)"
level=warn ts=2020-09-01T09:29:18.452Z caller=klog.go:86 component=k8s_client_runtime func=Warningf msg="/app/discovery/kubernetes/kubernetes.go:361: watch of *v1.Endpoints ended with: too old resource version: 77932955 (77936328)"
level=warn ts=2020-09-01T09:31:02.169Z caller=klog.go:86 component=k8s_client_runtime func=Warningf msg="/app/discovery/kubernetes/kubernetes.go:361: watch of *v1.Endpoints ended with: too old resource version: 77933801 (77937181)"
level=warn ts=2020-09-01T09:35:37.222Z caller=klog.go:86 component=k8s_client_runtime func=Warningf msg="/app/discovery/kubernetes/kubernetes.go:361: watch of *v1.Endpoints ended with: too old resource version: 77938858 (77939462)"
level=warn ts=2020-09-01T09:40:00.769Z caller=klog.go:86 component=k8s_client_runtime func=Warningf msg="/app/discovery/kubernetes/kubernetes.go:361: watch of *v1.Endpoints ended with: too old resource version: 77937615 (77941622)"
level=warn ts=2020-09-01T09:41:24.693Z caller=klog.go:86 component=k8s_client_runtime func=Warningf msg="/app/discovery/kubernetes/kubernetes.go:361: watch of *v1.Endpoints ended with: too old resource version: 77938126 (77942327)"
level=warn ts=2020-09-01T09:41:33.409Z caller=klog.go:86 component=k8s_client_runtime func=Warningf msg="/app/discovery/kubernetes/kubernetes.go:361: watch of *v1.Endpoints ended with: too old resource version: 77939775 (77942403)"
level=warn ts=2020-09-01T09:42:57.461Z caller=klog.go:86 component=k8s_client_runtime func=Warningf msg="/app/discovery/kubernetes/kubernetes.go:361: watch of *v1.Endpoints ended with: too old resource version: 77940485 (77943069)"
level=warn ts=2020-09-01T09:46:45.177Z caller=klog.go:86 component=k8s_client_runtime func=Warningf msg="/app/discovery/kubernetes/kubernetes.go:361: watch of *v1.Endpoints ended with: too old resource version: 77941349 (77944970)"
level=warn ts=2020-09-01T09:51:07.231Z caller=klog.go:86 component=k8s_client_runtime func=Warningf msg="/app/discovery/kubernetes/kubernetes.go:361: watch of *v1.Endpoints ended with: too old resource version: 77943616 (77947157)"
level=warn ts=2020-09-01T09:51:31.469Z caller=klog.go:86 component=k8s_client_runtime func=Warningf msg="/app/discovery/kubernetes/kubernetes.go:361: watch of *v1.Endpoints ended with: too old resource version: 77947266 (77947349)"
level=warn ts=2020-09-01T09:53:23.418Z caller=klog.go:86 component=k8s_client_runtime func=Warningf msg="/app/discovery/kubernetes/kubernetes.go:361: watch of *v1.Endpoints ended with: too old resource version: 77946542 (77948276)"
level=warn ts=2020-09-01T09:53:30.779Z caller=klog.go:86 component=k8s_client_runtime func=Warningf msg="/app/discovery/kubernetes/kubernetes.go:361: watch of *v1.Endpoints ended with: too old resource version: 77945788 (77948343)" ```

Answer 1

我将此作为社区 wiki 答案发布，以提高知名度。

我想了解导致错误的资源是什么？

正如@antoineco 提到的here

当您的集群中没有任何事情发生时，Kubernetes 客户端建立的手表没有机会自然刷新，并最终超时。这些消息只是表明这些手表正在重新创建。

正如@squat 提到的here。

这是 Kubernetes 中的一个已知事件，不是问题。 API 服务器会在它们非常旧时结束监视请求。 operator 使用 client-go informer，它负责自动重新列出资源，然后从最新的资源版本重新启动 watch。

---

我每秒看到它多次

根据您提供的日志，我不会说它每秒多次，每小时只有几次，如果您担心太多，我建议您在 prometheus-operator github 上询问。