Kubernetes 集群中 Ingress 故障的根因诊断云计算

作者：scwang18，主要负责技术架构，在容器云方向颇有研究。

前言 KubeSphere 是青云开源的基于 Kubernetes 的云原生分布式操作系统，提供了比较炫酷的 Kubernetes 集群管理界面，我们团队用 KubeSphere 来作为开发平台。
本文记录了一次 KubeSphere 环境下的网络故障的解决过程。
现象 【Kubernetes 集群中 Ingress 故障的根因诊断】开发同学反馈自己搭建的 Harbor 仓库总是出问题，偶尔会报 net/http: TLS handshake timeout ，通过 curl 的方式访问 harbor.xxxx.cn ，也会随机频繁挂起。但是 ping 的反馈一切正常。
原因分析接到错误报障后，经过了多轮分析，才最终定位到原因，应该是安装 KubeSphere 时，使用了最新版的 Kubernetes 1.23.1 。
虽然使用 ./kk version --show-supported-k8s 可以看到 KubeSphere 3.2.1 可以支持 Kubernetes 1.23.1 ，但实际上只是试验性支持，有坑的。
分析过程如下：

出现 Harbor registry 访问问题，下意识以为是 Harbor 部署有问题，但是在检查 Harbor core 的日志的时候，没有看到异常时有相应错误信息，甚至 info 级别的日志信息都没有。
又把目标放在 Harbor portal，查看访问日志，一样没有发现异常信息。
根据访问链，继续追查 kubesphere-router-kubesphere-system ，即 KubeSphere 版的 nginx ingress controller ，同样没有发现异常日志。
尝试在集群内其他 Pod 里访问 Harbor 的集群内 Service 地址，发现不会出现访问超时问题。初步判断是 KubeSphere 自带的 Ingress 的问题。
把 kubeSphere 自带的 Ingress Controller 关闭，安装 Kubernetes 官方推荐的 ingress-nginx-controller 版本，故障依旧，而且 Ingress 日志里也没有发现异常信息。
综合上面的分析，问题应该出现在客户端到 Ingress Controller 之间，我的 Ingress Controller 是通过 NodePort 方式暴露到集群外面。因此，测试其他通过 NodePort 暴露到集群外的 service，发现是一样的故障，至此，可以完全排除 Harbor 部署问题了，基本确定是客户端到 Ingress Controller 的问题。
外部客户端通过 NodePort 访问 Ingress Controller 时，会通过 kube-proxy 组件，分析 kube-proxy 的日志，发现告警信息

can’t set sysctl net/ipv4/vs/conn_reuse_mode, kernel version must be at least 4.1

这个告警信息是因为我的 centos 7.6 的内核版本过低，当前是 3.10.0-1160.21.1.el7.x86_64 ，与 Kubernetes 新版的 ipvs 存在兼容性问题。
可以通过升级操作系统的 kernel 版本可以解决。

升级完 kernel 后，Calico 启动不了，报以下错误信息

ipset v7.1: kernel and userspace incompatible: settype hash:ip,port with revision 6 not supported by userspace.

原因是安装 KubeSphere 时默认安装的 Calico 版本是 v3.20.0 , 这个版本不支持最新版的 Linux Kernel ，升级后的内核版本是 5.18.1-1.el7.elrepo.x86_64，calico 需要升级到 v3.23.0 以上版本。

升级完 Calico 版本后，Calico 继续报错

user "system:serviceaccount:kube-system:calico-node" cannot list resource "caliconodestatuses" in api group "crd.projectcalico.org"

还有另外一个错误信息，都是因为 clusterrole 的资源权限不足，可以通过修改 clusterrole 来解决问题。

至此，该莫名其妙的网络问题解决了。

解决过程根据上面的分析，主要解决方案如下：
升级操作系统内核

使用阿里云的 yum 源

wget -O /etc/yum.repos.d/CentOS-Base.repo http://mirrors.aliyun.com/repo/Centos-7.repo yum clean all && yum -y update

启用 elrepo 仓库

rpm --import https://www.elrepo.org/RPM-GPG-KEY-elrepo.org rpm -Uvh http://www.elrepo.org/elrepo-release-7.0-3.el7.elrepo.noarch.rpm

安装最新版本内核

yum --enablerepo=elrepo-kernel install kernel-ml

查看系统上的所有可用内核

awk -F\' '$1=="menuentry " {print i++ " : " $2}' /etc/grub2.cfg

设置新的内核为 grub2 的默认版本

查看第4步返回的系统可用内核列表，不出意外第1个应该是最新安装的内核。

grub2-set-default 0

生成 grub 配置文件并重启

grub2-mkconfig -o /boot/grub2/grub.cfg reboot now

验证

uname -r

升级 Calico
Kubernetes 上的 Calico 一般是使用 Daemonset 方式部署，我的集群里，Calico 的 Daemonset 名字是 calico-node。
直接输出为 yaml 文件，修改文件里的所有 image 版本号为最新版本 v3.23.1 。重新创建 Daemonset。

输出 yaml

kubectl -n kube-system get dscalico-node -o yaml>calico-node.yaml

calico-node.yaml：

apiVersion: apps/v1 kind: DaemonSet metadata: labels: k8s-app: calico-node name: calico-node namespace: kube-system spec: revisionHistoryLimit: 10 selector: matchLabels: k8s-app: calico-node template: metadata: creationTimestamp: null labels: k8s-app: calico-node spec: containers: - env: - name: DATASTORE_TYPE value: kubernetes - name: WAIT_FOR_DATASTORE value: "true" - name: NODENAME valueFrom: fieldRef: apiVersion: v1 fieldPath: spec.nodeName - name: CALICO_NETWORKING_BACKEND valueFrom: configMapKeyRef: key: calico_backend name: calico-config - name: CLUSTER_TYPE value: k8s,bgp - name: NODEIP valueFrom: fieldRef: apiVersion: v1 fieldPath: status.hostIP - name: IP_AUTODETECTION_METHOD value: can-reach=$(NODEIP) - name: IP value: autodetect - name: CALICO_IPV4POOL_IPIP value: Always - name: CALICO_IPV4POOL_VXLAN value: Never - name: FELIX_IPINIPMTU valueFrom: configMapKeyRef: key: veth_mtu name: calico-config - name: FELIX_VXLANMTU valueFrom: configMapKeyRef: key: veth_mtu name: calico-config - name: FELIX_WIREGUARDMTU valueFrom: configMapKeyRef: key: veth_mtu name: calico-config - name: CALICO_IPV4POOL_CIDR value: 10.233.64.0/18 - name: CALICO_IPV4POOL_BLOCK_SIZE value: "24" - name: CALICO_DISABLE_FILE_LOGGING value: "true" - name: FELIX_DEFAULTENDPOINTTOHOSTACTION value: ACCEPT - name: FELIX_IPV6SUPPORT value: "false" - name: FELIX_HEALTHENABLED value: "true" envFrom: - configMapRef: name: kubernetes-services-endpoint optional: true image: calico/node:v3.23.1 imagePullPolicy: IfNotPresent livenessProbe: exec: command: - /bin/calico-node - -felix-live - -bird-live failureThreshold: 6 initialDelaySeconds: 10 periodSeconds: 10 successThreshold: 1 timeoutSeconds: 10 name: calico-node readinessProbe: exec: command: - /bin/calico-node - -felix-ready - -bird-ready failureThreshold: 3 periodSeconds: 10 successThreshold: 1 timeoutSeconds: 10 resources: requests: cpu: 250m securityContext: privileged: true terminationMessagePath: /dev/termination-log terminationMessagePolicy: File volumeMounts: - mountPath: /host/etc/cni/net.d name: cni-net-dir - mountPath: /lib/modules name: lib-modules readOnly: true - mountPath: /run/xtables.lock name: xtables-lock - mountPath: /var/run/calico name: var-run-calico - mountPath: /var/lib/calico name: var-lib-calico - mountPath: /var/run/nodeagent name: policysync - mountPath: /sys/fs/ mountPropagation: Bidirectional name: sysfs - mountPath: /var/log/calico/cni name: cni-log-dir readOnly: true dnsPolicy: ClusterFirst hostNetwork: true initContainers: - command: - /opt/cni/bin/calico-ipam - -upgrade env: - name: KUBERNETES_NODE_NAME valueFrom: fieldRef: apiVersion: v1 fieldPath: spec.nodeName - name: CALICO_NETWORKING_BACKEND valueFrom: configMapKeyRef: key: calico_backend name: calico-config envFrom: - configMapRef: name: kubernetes-services-endpoint optional: true image: calico/cni:v3.23.1 imagePullPolicy: IfNotPresent name: upgrade-ipam resources: {} securityContext: privileged: true terminationMessagePath: /dev/termination-log terminationMessagePolicy: File volumeMounts: - mountPath: /var/lib/cni/networks name: host-local-net-dir - mountPath: /host/opt/cni/bin name: cni-bin-dir - command: - /opt/cni/bin/install env: - name: CNI_CONF_NAME value: 10-calico.conflist - name: CNI_NETWORK_CONFIG valueFrom: configMapKeyRef: key: cni_network_config name: calico-config - name: KUBERNETES_NODE_NAME valueFrom: fieldRef: apiVersion: v1 fieldPath: spec.nodeName - name: CNI_MTU valueFrom: configMapKeyRef: key: veth_mtu name: calico-config - name: SLEEP value: "false" envFrom: - configMapRef: name: kubernetes-services-endpoint optional: true image: calico/cni:v3.23.1 imagePullPolicy: IfNotPresent name: install-cni resources: {} securityContext: privileged: true terminationMessagePath: /dev/termination-log terminationMessagePolicy: File volumeMounts: - mountPath: /host/opt/cni/bin name: cni-bin-dir - mountPath: /host/etc/cni/net.d name: cni-net-dir - image: calico/pod2daemon-flexvol:v3.23.1 imagePullPolicy: IfNotPresent name: flexvol-driver resources: {} securityContext: privileged: true terminationMessagePath: /dev/termination-log terminationMessagePolicy: File volumeMounts: - mountPath: /host/driver name: flexvol-driver-host nodeSelector: kubernetes.io/os: linux priorityClassName: system-node-critical restartPolicy: Always schedulerName: default-scheduler securityContext: {} serviceAccount: calico-node serviceAccountName: calico-node terminationGracePeriodSeconds: 0 tolerations: - effect: NoSchedule operator: Exists - key: CriticalAddonsOnly operator: Exists - effect: NoExecute operator: Exists volumes: - hostPath: path: /lib/modules type: "" name: lib-modules - hostPath: path: /var/run/calico type: "" name: var-run-calico - hostPath: path: /var/lib/calico type: "" name: var-lib-calico - hostPath: path: /run/xtables.lock type: FileOrCreate name: xtables-lock - hostPath: path: /sys/fs/ type: DirectoryOrCreate name: sysfs - hostPath: path: /opt/cni/bin type: "" name: cni-bin-dir - hostPath: path: /etc/cni/net.d type: "" name: cni-net-dir - hostPath: path: /var/log/calico/cni type: "" name: cni-log-dir - hostPath: path: /var/lib/cni/networks type: "" name: host-local-net-dir - hostPath: path: /var/run/nodeagent type: DirectoryOrCreate name: policysync - hostPath: path: /usr/libexec/kubernetes/kubelet-plugins/volume/exec/nodeagent~uds type: DirectoryOrCreate name: flexvol-driver-host updateStrategy: rollingUpdate: maxSurge: 0 maxUnavailable: 1 type: RollingUpdate

ClusterRole
还需要修改 ClusterRole ，否则 Calico 会一直报权限错。

输出 yaml

kubectl get clusterrole calico-node -o yaml >calico-node-clusterrole.yaml

calico-node-clusterrole.yaml

apiVersion: rbac.authorization.k8s.io/v1 kind: ClusterRole metadata: name: calico-node rules: - apiGroups: - "" resources: - pods - nodes - namespaces verbs: - get - apiGroups: - discovery.k8s.io resources: - endpointslices verbs: - watch - list - apiGroups: - "" resources: - endpoints - services verbs: - watch - list - get - apiGroups: - "" resources: - configmaps verbs: - get - apiGroups: - "" resources: - nodes/status verbs: - patch - update - apiGroups: - networking.k8s.io resources: - networkpolicies verbs: - watch - list - apiGroups: - "" resources: - pods - namespaces - serviceaccounts verbs: - list - watch - apiGroups: - "" resources: - pods/status verbs: - patch - apiGroups: - crd.projectcalico.org resources: - globalfelixconfigs - felixconfigurations - bgppeers - globalbgpconfigs - bgpconfigurations - ippools - ipamblocks - globalnetworkpolicies - globalnetworksets - networkpolicies - networksets - clusterinformations - hostendpoints - blockaffinities - caliconodestatuses - ipreservations verbs: - get - list - watch - apiGroups: - crd.projectcalico.org resources: - ippools - felixconfigurations - clusterinformations verbs: - create - update - apiGroups: - "" resources: - nodes verbs: - get - list - watch - apiGroups: - crd.projectcalico.org resources: - bgpconfigurations - bgppeers verbs: - create - update - apiGroups: - crd.projectcalico.org resources: - blockaffinities - ipamblocks - ipamhandles verbs: - get - list - create - update - delete - apiGroups: - crd.projectcalico.org resources: - ipamconfigs verbs: - get - apiGroups: - crd.projectcalico.org resources: - blockaffinities verbs: - watch - apiGroups: - apps resources: - daemonsets verbs: - get

总结这次奇怪的网络故障，最终原因还是因为 KubeSphere 的版本与 Kubernetes 的版本不匹配。所以工作环境要稳字为先，不要冒进使用最新的版本。否则会耽搁很多时间来解决莫名其妙的问题。