[k8s]k8s 1.9(on the fly搭建) 1.9_cni-flannel部署排错 ipvs模式
角色 | 节点名 | 节点ip |
---|---|---|
master | n1 | 192.168.14.11 |
节点1 | n2 | 192.168.14.12 |
节点2 | n3 | 192.168.14.13 |
https://raw.githubusercontent.com/lannyMa/scripts/master/k8s/
准备token.csv和bootstrap.kubeconfig文件
- 在master生成token.csv
BOOTSTRAP_TOKEN="41f7e4ba8b7be874fcff18bf5cf41a7c"
cat > token.csv<<EOF
41f7e4ba8b7be874fcff18bf5cf41a7c,kubelet-bootstrap,10001,"system:kubelet-bootstrap"
EOF
- 将bootstrap.kubeconfig同步到所有节点
设置集群参数
kubectl config set-cluster kubernetes \
--certificate-authority=/root/ssl/ca.crt \
--embed-certs=true \
--server=http://192.168.14.11:8080 \
--kubeconfig=bootstrap.kubeconfig
设置客户端认证参数
kubectl config set-credentials kubelet-bootstrap \
--token="41f7e4ba8b7be874fcff18bf5cf41a7c" \
--kubeconfig=bootstrap.kubeconfig
设置上下文参数
kubectl config set-context default \
--cluster=kubernetes \
--user=kubelet-bootstrap \
--kubeconfig=bootstrap.kubeconfig
设置默认上下文
kubectl config use-context default --kubeconfig=bootstrap.kubeconfig
两个文件我都放在了/root下.
coredns+dashboard(heapster)+kube-router yaml
https://github.com/lannyMa/scripts/tree/master/k8s
k8s 1.9 on the fly启动
etcd --advertise-client-urls=http://192.168.14.11:2379 --listen-client-urls=http://0.0.0.0:2379 --debug
kube-apiserver --service-cluster-ip-range=10.254.0.0/16 --etcd-servers=http://127.0.0.1:2379 --insecure-bind-address=0.0.0.0 --admission-control=ServiceAccount --service-account-key-file=/root/ssl/ca.key --client-ca-file=/root/ssl/ca.crt --tls-cert-file=/root/ssl/server.crt --tls-private-key-file=/root/ssl/server.key --allow-privileged=true --storage-backend=etcd2 --v=2 --enable-bootstrap-token-auth --token-auth-file=/root/token.csv
kube-controller-manager --master=http://127.0.0.1:8080 --service-account-private-key-file=/root/ssl/ca.key --cluster-signing-cert-file=/root/ssl/ca.crt --cluster-signing-key-file=/root/ssl/ca.key --root-ca-file=/root/ssl/ca.crt --v=2
kube-scheduler --master=http://127.0.0.1:8080 --v=2
kubelet --allow-privileged=true --cluster-dns=10.254.0.2 --cluster-domain=cluster.local --v=2 --experimental-bootstrap-kubeconfig=/root/bootstrap.kubeconfig --kubeconfig=/root/kubelet.kubeconfig --fail-swap-on=false
kube-proxy --master=http://192.168.14.11:8080 --v=2
kubectl get csr | grep Pending | awk '{print $1}' | xargs kubectl certificate approve
api相对1.7的变化:
https://github.com/kubernetes/kubernetes/blob/master/CHANGELOG.md#before-upgrading
https://mritd.me/2017/10/09/set-up-kubernetes-1.8-ha-cluster/
- kubelet没了--api-servers参数,必须用bootstrap方式去连api
- 移除了 --runtime-config=rbac.authorization.k8s.io/v1beta1 配置,因为 RBAC 已经稳定,被纳入了 v1 api,不再需要指定开启
- --authorization-mode 授权模型增加了 Node 参数,因为 1.8 后默认 system:node role 不会自动授予 system:nodes 组
- 增加 --audit-policy-file 参数用于指定高级审计配置
- 移除 --experimental-bootstrap-token-auth 参数,更换为 --enable-bootstrap-token-auth
k8s1.9 cni(flannel) on the fly启动
注: HostPort不能使用CNI网络插件(docker run -p 8081:8080)。这意味着pod中所有HostPort属性将被简单地忽略。
mkdir -p /etc/cni/net.d /opt/cni/bin
wget https://github.com/containernetworking/plugins/releases/download/v0.6.0/cni-plugins-amd64-v0.6.0.tgz
tar xf cni-plugins-amd64-v0.6.0.tgz -C /opt/cni/bin
cat > /etc/cni/net.d/10-flannel.conflist<<EOF
{
"name":"cni0",
"cniVersion":"0.3.1",
"plugins":[
{
"type":"flannel",
"delegate":{
"forceAddress":true,
"isDefaultGateway":true
}
},
{
"type":"portmap",
"capabilities":{
"portMappings":true
}
}
]
}
EOF
etcd --advertise-client-urls=http://192.168.14.11:2379 --listen-client-urls=http://0.0.0.0:2379 --debug
kube-apiserver --service-cluster-ip-range=10.254.0.0/16 --etcd-servers=http://127.0.0.1:2379 --insecure-bind-address=0.0.0.0 --admission-control=ServiceAccount --service-account-key-file=/root/ssl/ca.key --client-ca-file=/root/ssl/ca.crt --tls-cert-file=/root/ssl/server.crt --tls-private-key-file=/root/ssl/server.key --allow-privileged=true --storage-backend=etcd2 --v=2 --enable-bootstrap-token-auth --token-auth-file=/root/token.csv
kube-controller-manager --master=http://127.0.0.1:8080 --service-account-private-key-file=/root/ssl/ca.key --cluster-signing-cert-file=/root/ssl/ca.crt --cluster-signing-key-file=/root/ssl/ca.key --root-ca-file=/root/ssl/ca.crt --v=2 --allocate-node-cidrs=true --cluster-cidr=10.244.0.0/16
kube-scheduler --master=http://127.0.0.1:8080 --v=2
kubelet --allow-privileged=true --cluster-dns=10.254.0.2 --cluster-domain=cluster.local --v=2 --experimental-bootstrap-kubeconfig=/root/bootstrap.kubeconfig --kubeconfig=/root/kubelet.kubeconfig --fail-swap-on=false --network-plugin=cni
kube-proxy --master=http://192.168.14.11:8080 --v=2
kubectl apply -f https://raw.githubusercontent.com/lannyMa/scripts/master/k8s/cni-flannel.yml
查看pod确实从cni0分到了地址
[root@n1 ~]# kk
NAMESPACE NAME READY STATUS RESTARTS AGE IP NODE LABELS
default b1 1/1 Running 0 6m 10.244.0.2 n2.ma.com <none>
default b2 1/1 Running 0 6m 10.244.1.2 n3.ma.com <none>
[root@n1 ~]# kubectl exec -it b1 sh
/ # ping 10.244.1.2
PING 10.244.1.2 (10.244.1.2): 56 data bytes
64 bytes from 10.244.1.2: seq=0 ttl=62 time=6.292 ms
64 bytes from 10.244.1.2: seq=1 ttl=62 time=0.981 ms
遇到的报错
kubectl apply -f cni-flannel.yml没提示报错,但get pod无显示
kubectl apply -f https://raw.githubusercontent.com/lannyMa/scripts/master/k8s/cni-flannel.yml
原因: yaml用到了sa未创建
---
apiVersion: v1
kind: ServiceAccount
metadata:
name: flannel
namespace: kube-system
pod0无cni0网卡,创建pod后分到的是172.17.x.x即docker0分配的地址
原因: kubelet未加cni启动参数
--network-plugin=cni
创建成功后提示.kubectl create成功后一直pending,kubelet报错
因为ipv6没开,参考how-to-disable-ipv6
ifconfig -a | grep inet6
I1231 23:22:08.343188 15369 kubelet.go:1881] SyncLoop (PLEG): "busybox_default(5a16fa0c-ee3e-11e7-9220-000c29bfdc52)", event: &pleg.PodLifecycleEvent{ID:"5a16fa0c-ee3e-11e7-9220-000c29bfdc52", Type:"ContainerDied", Data:"76e584c83f8dd3d54c759ac637bb47aa2a94de524372e282abde004d1cfbcd1b"}
W1231 23:22:08.343338 15369 pod_container_deletor.go:77] Container "76e584c83f8dd3d54c759ac637bb47aa2a94de524372e282abde004d1cfbcd1b" not found in pod's containers
I1231 23:22:08.644401 15369 kuberuntime_manager.go:403] No ready sandbox for pod "busybox_default(5a16fa0c-ee3e-11e7-9220-000c29bfdc52)" can be found. Need to start a new one
E1231 23:22:08.857121 15369 cni.go:259] Error adding network: open /proc/sys/net/ipv6/conf/eth0/accept_dad: no such file or directory
E1231 23:22:08.857144 15369 cni.go:227] Error while adding to cni network: open /proc/sys/net/ipv6/conf/eth0/accept_dad: no such file or directory
E1231 23:22:08.930343 15369 remote_runtime.go:92] RunPodSandbox from runtime service failed: rpc error: code = Unknown desc = NetworkPlugin cni failed to set up pod "busybox_default" network: open /proc/sys/net/ipv6/conf/eth0/accept_dad: no such file or directory
多次实验,导致node的网络配置污染,所以新验证时一定要reboot,清理环境
rm -rf /var/lib/kubelet/
docker rm $(docker ps -a -q)
controller需要加这两个参数,且cluster-cidr的地址和cni-flannel.yaml里的地址要一致.和svc网段的地址(--service-cluster-ip-range)不要相同.
--allocate-node-cidrs=true --cluster-cidr=10.244.0.0/16
no IP addresses available in range set: 10.244.0.1-10.244.0.254
没地址了,pod一直在创建中....
参考: https://github.com/kubernetes/kubernetes/issues/57280
现象:
- kubelet报错
E0101 00:06:38.629105 1109 kuberuntime_manager.go:647] createPodSandbox for pod "busybox2_default(7fa06467-ee44-11e7-a440-000c29bfdc52)" failed: rpc error: code = Unknown desc = NetworkPlugin cni failed to set up pod "busybox2_default" network: failed to allocate for range 0: no IP addresses available in range set: 10.244.0.1-10.244.0.254
E0101 00:06:38.629143 1109 pod_workers.go:186] Error syncing pod 7fa06467-ee44-11e7-a440-000c29bfdc52 ("busybox2_default(7fa06467-ee44-11e7-a440-000c29bfdc52)"), skipping: failed to "CreatePodSandbox" for "busybox2_default(7fa06467-ee44-11e7-a440-000c29bfdc52)" with CreatePodSandboxError: "CreatePodSandbox for pod \"busybox2_default(7fa06467-ee44-11e7-a440-000c29bfdc52)\" failed: rpc error: code = Unknown desc = NetworkPlugin cni failed to set up pod \"busybox2_default\" network: failed to allocate for range 0: no IP addresses available in range set: 10.244.0.1-10.244.0.254"
- ip地址
/var/lib/cni/networks# ls cbr0/
10.244.0.10 10.244.0.123 10.244.0.147 10.244.0.170 10.244.0.194 10.244.0.217 10.244.0.240 10.244.0.35 10.244.0.59 10.244.0.82
10.244.0.100 10.244.0.124 10.244.0.148 10.244.0.171 10.244.0.195 10.244.0.218 10.244.0.241 10.244.0.36 10.244.0.6 10.244.0.83
10.244.0.101 10.244.0.125 10.244.0.149 10.244.0.172 10.244.0.196 10.244.0.219 10.244.0.242 10.244.0.37 10.244.0.60 10.244.0.84
10.244.0.102 10.244.0.126 10.244.0.15 10.244.0.173 10.244.0.197 10.244.0.22 10.244.0.243 10.244.0.38 10.244.0.61 10.244.0.85
10.244.0.103 10.244.0.127 10.244.0.150 10.244.0.174 10.244.0.198 10.244.0.220 10.244.0.244 10.244.0.39 10.244.0.62 10.244.0.86
10.244.0.104 10.244.0.128 10.244.0.151 10.244.0.175 10.244.0.199 10.244.0.221 10.244.0.245 10.244.0.4 10.244.0.63 10.244.0.87
10.244.0.105 10.244.0.129 10.244.0.152 10.244.0.176 10.244.0.2 10.244.0.222 10.244.0.246 10.244.0.40 10.244.0.64 10.244.0.88
10.244.0.106 10.244.0.13 10.244.0.153 10.244.0.177 10.244.0.20 10.244.0.223 10.244.0.247 10.244.0.41 10.244.0.65 10.244.0.89
10.244.0.107 10.244.0.130 10.244.0.154 10.244.0.178 10.244.0.200 10.244.0.224 10.244.0.248 10.244.0.42 10.244.0.66 10.244.0.9
10.244.0.108 10.244.0.131 10.244.0.155 10.244.0.179 10.244.0.201 10.244.0.225 10.244.0.249 10.244.0.43 10.244.0.67 10.244.0.90
10.244.0.109 10.244.0.132 10.244.0.156 10.244.0.18 10.244.0.202 10.244.0.226 10.244.0.25 10.244.0.44 10.244.0.68 10.244.0.91
10.244.0.11 10.244.0.133 10.244.0.157 10.244.0.180 10.244.0.203 10.244.0.227 10.244.0.250 10.244.0.45 10.244.0.69 10.244.0.92
10.244.0.110 10.244.0.134 10.244.0.158 10.244.0.181 10.244.0.204 10.244.0.228 10.244.0.251 10.244.0.46 10.244.0.7 10.244.0.93
10.244.0.111 10.244.0.135 10.244.0.159 10.244.0.182 10.244.0.205 10.244.0.229 10.244.0.252 10.244.0.47 10.244.0.70 10.244.0.94
10.244.0.112 10.244.0.136 10.244.0.16 10.244.0.183 10.244.0.206 10.244.0.23 10.244.0.253 10.244.0.48 10.244.0.71 10.244.0.95
10.244.0.113 10.244.0.137 10.244.0.160 10.244.0.184 10.244.0.207 10.244.0.230 10.244.0.254 10.244.0.49 10.244.0.72 10.244.0.96
10.244.0.114 10.244.0.138 10.244.0.161 10.244.0.185 10.244.0.208 10.244.0.231 10.244.0.26 10.244.0.5 10.244.0.73 10.244.0.97
10.244.0.115 10.244.0.139 10.244.0.162 10.244.0.186 10.244.0.209 10.244.0.232 10.244.0.27 10.244.0.50 10.244.0.74 10.244.0.98
10.244.0.116 10.244.0.14 10.244.0.163 10.244.0.187 10.244.0.21 10.244.0.233 10.244.0.28 10.244.0.51 10.244.0.75 10.244.0.99
10.244.0.117 10.244.0.140 10.244.0.164 10.244.0.188 10.244.0.210 10.244.0.234 10.244.0.29 10.244.0.52 10.244.0.76 last_reserved_ip.0
10.244.0.118 10.244.0.141 10.244.0.165 10.244.0.189 10.244.0.211 10.244.0.235 10.244.0.3 10.244.0.53 10.244.0.77
10.244.0.119 10.244.0.142 10.244.0.166 10.244.0.19 10.244.0.212 10.244.0.236 10.244.0.30 10.244.0.54 10.244.0.78
10.244.0.12 10.244.0.143 10.244.0.167 10.244.0.190 10.244.0.213 10.244.0.237 10.244.0.31 10.244.0.55 10.244.0.79
10.244.0.120 10.244.0.144 10.244.0.168 10.244.0.191 10.244.0.214 10.244.0.238 10.244.0.32 10.244.0.56 10.244.0.8
10.244.0.121 10.244.0.145 10.244.0.169 10.244.0.192 10.244.0.215 10.244.0.239 10.244.0.33 10.244.0.57 10.244.0.80
10.244.0.122 10.244.0.146 10.244.0.17 10.244.0.193 10.244.0.216 10.244.0.24 10.244.0.34 10.244.0.58 10.244.0.81
- flannel创建了很多文件
/var/lib/cni/flannel# ls | wc ; date
解决:
干掉cni-flannel,停运集群.清理环境.
rm -rf /var/lib/cni/flannel/* && rm -rf /var/lib/cni/networks/cbr0/* && ip link delete cni0
rm -rf /var/lib/cni/networks/cni0/*
- 推荐打开,不打开我没发现什么问题
echo 'net.bridge.bridge-nf-call-iptables=1' >> /etc/sysctl.conf
sysctl -p
# 打开ip转发,下面4条都干上去
net.ipv4.ip_forward = 1
# Disable netfilter on bridges.
net.bridge.bridge-nf-call-iptables = 1
net.bridge.bridge-nf-call-ip6tables = 1
net.bridge.bridge-nf-call-arptables = 1
参考:
https://raw.githubusercontent.com/coreos/flannel/master/Documentation/kube-flannel.yml
http://cizixs.com/2017/05/23/container-network-cni
https://k8smeetup.github.io/docs/concepts/cluster-administration/network-plugins/
https://mritd.me/2017/09/20/set-up-ha-kubernetes-cluster-on-aliyun-ecs/
https://coreos.com/flannel/docs/latest/kubernetes.html (不要用legency的那个,那个是kubelet模式)
https://feisky.gitbooks.io/kubernetes/network/flannel/#cni集成
http://blog.csdn.net/idea77/article/details/78793318
kube-proxy ipvs模式
目前还是测试版,打开玩一玩.
参考: https://jicki.me/2017/12/20/kubernetes-1.9-ipvs/#启动-kube-proxy
https://mritd.me/2017/10/10/kube-proxy-use-ipvs-on-kubernetes-1.8/
确保内核有rr模块
[root@n2 ~]# lsmod | grep ip_vs
ip_vs_sh 12688 0
ip_vs_wrr 12697 0
ip_vs_rr 12600 2
ip_vs 141092 8 ip_vs_rr,ip_vs_sh,ip_vs_wrr
nf_conntrack 133387 9 ip_vs,nf_nat,nf_nat_ipv4,nf_nat_ipv6,xt_conntrack,nf_nat_masquerade_ipv4,nf_conntrack_netlink,nf_conntrack_ipv4,nf_conntrack_ipv6
libcrc32c 12644 4 xfs,ip_vs,nf_nat,nf_conntrack
启用 ipvs 后与 1.7 版本的配置差异如下:
增加 --feature-gates=SupportIPVSProxyMode=true 选项,用于告诉 kube-proxy 开启 ipvs 支持,因为目前 ipvs 并未稳定
增加 ipvs-min-sync-period、--ipvs-sync-period、--ipvs-scheduler 三个参数用于调整 ipvs,具体参数值请自行查阅 ipvs 文档
增加 --masquerade-all 选项,以确保反向流量通过
重点说一下 --masquerade-all 选项: kube-proxy ipvs 是基于 NAT 实现的,当创建一个 service 后,kubernetes 会在每个节点上创建一个网卡,同时帮你将 Service IP(VIP) 绑定上,此时相当于每个 Node 都是一个 ds,而其他任何 Node 上的 Pod,甚至是宿主机服务(比如 kube-apiserver 的 6443)都可能成为 rs;按照正常的 lvs nat 模型,所有 rs 应该将 ds 设置成为默认网关,以便数据包在返回时能被 ds 正确修改;在 kubernetes 将 vip 设置到每个 Node 后,默认路由显然不可行,所以要设置 --masquerade-all 选项,以便反向数据包能通过
注意:--masquerade-all 选项与 Calico 安全策略控制不兼容,请酌情使用
kube-proxy --master=http://192.168.14.11:8080 --v=2 --feature-gates=SupportIPVSProxyMode=true --masquerade-all --proxy-mode=ipvs --masquerade-all
注意:
1.需要打开 –feature-gates=SupportIPVSProxyMode=true,官方 –feature-gates=SupportIPVSProxyMode=false 默认是 false
2.–masquerade-all 必须添加这项配置,否则 创建 svc 在 ipvs 不会添加规则
3.打开 ipvs 需要安装 ipvsadm 软件, 在 node 中安装
yum install ipvsadm -y
ipvsadm -L -n
[root@n2 ~]# ipvsadm -L -n
IP Virtual Server version 1.2.1 (size=4096)
Prot LocalAddress:Port Scheduler Flags
-> RemoteAddress:Port Forward Weight ActiveConn InActConn
TCP 10.254.0.1:443 rr persistent 10800
-> 192.168.14.11:6443 Masq 1 0 0
TCP 10.254.12.188:80 rr
-> 10.244.0.3:80 Masq 1 0 0
-> 10.244.1.4:80 Masq 1 0 0