calico部署与网卡绑定

calico问题排障

这个问题几乎每个人都会遇到。因为官方的step by step太傻白甜，没有把IP_AUTODETECTION_METHOD这个IP检测方法的参数放入calico.yaml中，calico会使用第一个找到的network interface（往往是错误的interface），导致Calico把master也算进nodes，于是master BGP启动失败，而其他workers则启动成功。

问题描述

k8s集群安装网络组件calico后，查看pod

[ansible@k8s-cp calico]$ kubectl get pod --all-namespaces
NAMESPACE     NAME                             READY   STATUS    RESTARTS   AGE
kube-system   calico-node-jm74b                1/2     Running   0          7m16s
kube-system   calico-node-xk4fg                1/2     Running   0          2m5s
kube-system   coredns-7b47b4c577-447cn         1/1     Running   0          8m27s
kube-system   coredns-7b47b4c577-svm5v         1/1     Running   0          8m27s
kube-system   etcd-k8s-cp                      1/1     Running   0          7m51s
kube-system   kube-apiserver-k8s-cp            1/1     Running   0          8m1s
kube-system   kube-controller-manager-k8s-cp   1/1     Running   0          8m4s
kube-system   kube-proxy-nzmhh                 1/1     Running   0          8m27s
kube-system   kube-proxy-pjbp8                 1/1     Running   0          2m5s
kube-system   kube-scheduler-k8s-cp            1/1     Running   0          7m43s

等待几分钟后，pod calico-node-jm74b和calico-node-xk4fg的READY值依然是1/2

查看pod calico-node-xk4fg的详细信息，发现有如下错误信息

Warning  Unhealthy  11s (x19 over 3m11s)  kubelet, k8s-agent-1  Readiness probe failed: calico/node is not ready: BIRD is not ready: BGP not established with 172.18.0.1

这个问题会导致部署完业务容器后，网络上有问题，导致业务无法正常访问。

问题解决

calico在多网络接口时自动检测到错误的网络接口,导致网络无法连通，通过指定网络接口(网卡名)解决问题

修改calico.yaml

在

- name: CLUSTER_TYPE
  value: "k8s,bgp" 

下增加两行

- name: IP_AUTODETECTION_METHOD
  value: "interface=enp0s3"

enp0s3是我机器的网卡名

重新部署网络组件calico, READY值变为2/2

[ansible@k8s-cp calico]$ kubectl get pod --all-namespaces
NAMESPACE     NAME                             READY   STATUS    RESTARTS   AGE
kube-system   calico-node-jm74b                2/2     Running   0          15m
kube-system   calico-node-xk4fg                2/2     Running   0          9m51s
kube-system   coredns-7b47b4c577-447cn         1/1     Running   0          16m
kube-system   coredns-7b47b4c577-svm5v         1/1     Running   0          16m
kube-system   etcd-k8s-cp                      1/1     Running   0          15m
kube-system   kube-apiserver-k8s-cp            1/1     Running   0          15m
kube-system   kube-controller-manager-k8s-cp   1/1     Running   0          15m
kube-system   kube-proxy-nzmhh                 1/1     Running   0          16m
kube-system   kube-proxy-pjbp8                 1/1     Running   0          9m51s
kube-system   kube-proxy-wgz2c                 1/1     Running   0          114s
kube-system   kube-scheduler-k8s-cp            1/1     Running   0          15m