我在 AWS 云上使用 KOPS 创建了一个 Kubernetes 集群。集群创建时没有任何问题，可以正常运行 10-15 小时。我已经在这个集群上部署了 SAP Vora2.1。然而，一般在 12-15 小时后，KOPS 集群会遇到与 kube-proxy 和 kube-dns 相关的问题。这些 pod 要么关闭，要么显示为已完成状态。还有很多重启。这最终导致我的应用程序 pod 出现问题，并且应用程序也出现故障。该应用程序使用 consul 进行服务发现，但是由于 kubernetes 基础服务无法正常工作，因此即使我尝试恢复 kube-proxy/kube-dns pod，应用程序也不会进入稳定状态。
这是一个以完全自动缩放模式设置的 3 节点集群（1 个主节点和 2 个节点）。覆盖网络使用默认的 kubenet。以下是系统遇到问题后 pod 状态的快照，
[root@ip-172-31-18-162 ~]# kubectl get pods --all-namespaces
NAMESPACE NAME READY STATUS RESTARTS AGE
infyvora vora-catalog-1549734119-cfnhz 0/2 CrashLoopBackOff 188 20h
infyvora vora-consul-0 0/1 CrashLoopBackOff 101 20h
infyvora vora-consul-1 1/1 Running 34 20h
infyvora vora-consul-2 0/1 CrashLoopBackOff 95 20h
infyvora vora-deployment-operator-293895365-4b3t6 0/1 Completed 104 20h
infyvora vora-disk-0 1/2 CrashLoopBackOff 187 20h
infyvora vora-dlog-0 0/2 CrashLoopBackOff 226 20h
infyvora vora-dlog-1 1/2 CrashLoopBackOff 155 20h
infyvora vora-doc-store-2451237348-dkrm6 0/2 CrashLoopBackOff 229 20h
infyvora vora-elasticsearch-logging-v1-444540252-mwfrz 0/1 CrashLoopBackOff 100 20h
infyvora vora-elasticsearch-logging-v1-444540252-vrr63 1/1 Running 14 20h
infyvora vora-elasticsearch-retention-policy-137762458-ns5pc 1/1 Running 13 20h
infyvora vora-fluentd-kubernetes-v1.21-9f4pt 1/1 Running 12 20h
infyvora vora-fluentd-kubernetes-v1.21-s2t1j 0/1 CrashLoopBackOff 99 20h
infyvora vora-grafana-2929546178-vrf5h 1/1 Running 13 20h
infyvora vora-graph-435594712-47lcg 0/2 CrashLoopBackOff 157 20h
infyvora vora-kibana-logging-3693794794-2qn86 0/1 CrashLoopBackOff 99 20h
infyvora vora-landscape-2532068267-w1f5n 0/2 CrashLoopBackOff 232 20h
infyvora vora-nats-streaming-1569990702-kcl1v 1/1 Running 13 20h
infyvora vora-prometheus-node-exporter-k4c3g 0/1 CrashLoopBackOff 102 20h
infyvora vora-prometheus-node-exporter-xp511 1/1 Running 13 20h
infyvora vora-prometheus-pushgateway-399610745-tcfk7 0/1 CrashLoopBackOff 103 20h
infyvora vora-prometheus-server-3955170982-xpct0 2/2 Running 24 20h
infyvora vora-relational-376953862-w39tc 0/2 CrashLoopBackOff 237 20h
infyvora vora-security-operator-2514524099-7ld0k 0/1 CrashLoopBackOff 103 20h
infyvora vora-thriftserver-409431919-8c1x9 2/2 Running 28 20h
infyvora vora-time-series-1188816986-f2fbq 1/2 CrashLoopBackOff 184 20h
infyvora vora-tools5tlpt-100252330-mrr9k 0/1 rpc error: code = 4 desc = context deadline exceeded 272 17h
infyvora vora-tools6zr3m-3592177467-n7sxd 0/1 Completed 1 20h
infyvora vora-tx-broker-4168728922-hf8jz 0/2 CrashLoopBackOff 151 20h
infyvora vora-tx-coordinator-3910571185-l0r4n 0/2 CrashLoopBackOff 184 20h
infyvora vora-tx-lock-manager-2734670982-bn7kk 0/2 Completed 228 20h
infyvora vsystem-1230763370-5ckr0 0/1 CrashLoopBackOff 115 20h
infyvora vsystem-auth-1068224543-0g59w 0/1 CrashLoopBackOff 102 20h
infyvora vsystem-vrep-1427606801-zprlr 0/1 CrashLoopBackOff 121 20h
kube-system dns-controller-3110272648-chwrs 1/1 Running 0 22h
kube-system etcd-server-events-ip-172-31-64-102.ap-southeast-1.compute.internal 1/1 Running 0 22h
kube-system etcd-server-ip-172-31-64-102.ap-southeast-1.compute.internal 1/1 Running 0 22h
kube-system kube-apiserver-ip-172-31-64-102.ap-southeast-1.compute.internal 1/1 Running 0 22h
kube-system kube-controller-manager-ip-172-31-64-102.ap-southeast-1.compute.internal 1/1 Running 0 22h
kube-system kube-dns-1311260920-cm1fs 0/3 Completed 309 22h
kube-system kube-dns-1311260920-hm5zd 3/3 Running 39 22h
kube-system kube-dns-autoscaler-1818915203-wmztj 1/1 Running 12 22h
kube-system kube-proxy-ip-172-31-64-102.ap-southeast-1.compute.internal 1/1 Running 0 22h
kube-system kube-proxy-ip-172-31-64-110.ap-southeast-1.compute.internal 0/1 CrashLoopBackOff 98 22h
kube-system kube-proxy-ip-172-31-64-15.ap-southeast-1.compute.internal 1/1 Running 13 22h
kube-system kube-scheduler-ip-172-31-64-102.ap-southeast-1.compute.internal 1/1 Running 0 22h
kube-system tiller-deploy-352283156-97hhb 1/1 Running 34 22h
有没有人在 AWS 上遇到过与 KOPS kubernetes 相关的类似问题。欣赏是否有解决此问题的指针。
问候， 迪帕克