使用kubeadm部署Kubernetes 1.29
2023-12-27
kubeadm是Kubernetes官方提供的用于快速安部署Kubernetes集群的工具。伴随Kubernetes每个版本的发布，kubeadm可能会对集群配置方面的一些实践做调整，通过实验kubeadm可以学习到Kubernetes官方在集群配置上一些新的最佳实践。
1.准备 #
1.1 系统配置 #
在安装之前，需要先做好如下准备。
3台Linux主机如下：
- node4 - Ubuntu 22.04
- node5 - openEuler release 22.03 (LTS-SP2)
- node6 - Rocky Linux release 8.8 (Green Obsidian)
1cat /etc/hosts
2192.168.96.154 node4
3192.168.96.155 node5
4192.168.96.156 node6
在各个主机上完成下面的系统配置。
如果系统启用了selinux，使用下面的命令禁用selinux:
1setenforce 0
2
3vi /etc/selinux/config
4SELINUX=disabled
如果各个主机启用了防火墙策略，需要开放Kubernetes各个组件所需要的端口，可以查看Ports and Protocols中的内容, 开放相关端口或者关闭主机的防火墙。
创建/etc/modules-load.d/containerd.conf配置文件，确保在系统启动时自动加载所需的内核模块，以满足容器运行时的要求:
1cat << EOF > /etc/modules-load.d/containerd.conf
2overlay
3br_netfilter
4EOF
执行以下命令使配置生效:
1modprobe overlay
2modprobe br_netfilter
创建/etc/sysctl.d/99-kubernetes-cri.conf配置文件：
1cat << EOF > /etc/sysctl.d/99-kubernetes-cri.conf
2net.bridge.bridge-nf-call-ip6tables = 1
3net.bridge.bridge-nf-call-iptables = 1
4net.ipv4.ip_forward = 1
5user.max_user_namespaces=28633
6EOF
执行以下命令使配置生效:
1sysctl -p /etc/sysctl.d/99-kubernetes-cri.conf
在文件名
/etc/sysctl.d/99-kubernetes-cri.conf中，“99” 代表文件的优先级或顺序。sysctl是Linux内核参数的配置工具，它可以通过修改/proc/sys/目录下的文件来设置内核参数。在/etc/sysctl.d/目录中，可以放置一系列的配置文件，以便在系统启动时自动加载这些参数。这些配置文件按照文件名的字母顺序逐个加载。数字前缀用于指定加载的顺序，较小的数字表示较高的优先级。
1.2 配置服务器支持开启ipvs的前提条件 #
由于ipvs已经加入到了内核的主干，所以为kube-proxy开启ipvs的前提需要加载以下的内核模块：
1ip_vs
2ip_vs_rr
3ip_vs_wrr
4ip_vs_sh
5nf_conntrack_ipv4
创建/etc/modules-load.d/ipvs.conf文件，保证在节点重启后能自动加载所需模块:
1cat > /etc/modules-load.d/ipvs.conf <<EOF
2ip_vs
3ip_vs_rr
4ip_vs_wrr
5ip_vs_sh
6EOF
执行以下命令使配置立即生效:
1modprobe ip_vs
2modprobe ip_vs_rr
3modprobe ip_vs_wrr
4modprobe ip_vs_sh
使用lsmod | grep -e ip_vs -e nf_conntrack命令查看是否已经正确加载所需的内核模块。
接下来还需要确保各个节点上已经安装了ipset软件包，为了便于查看ipvs的代理规则，最好安装一下管理工具ipvsadm。
在Ubuntu系统上执行:
1apt install -y ipset ipvsadm
在openEuler或Rocky Linux系统上执行:
1yum install -y ipset ipvsadm
如果不满足以上前提条件，即使kube-proxy的配置开启了ipvs模式，也会退回到iptables模式。
1.3 部署容器运行时Containerd #
在各个服务器节点上安装容器运行时Containerd。
下载Containerd的二进制包， 需要注意cri-containerd-(cni-)-VERSION-OS-ARCH.tar.gz发行包自containerd 1.6版本起已经被弃用，在某些 Linux 发行版上无法正常工作，并将在containerd 2.0版本中移除，这里下载containerd-<VERSION>-<OS>-<ARCH>.tar.gz的发行包，后边再单独下载安装runc和CNI plugins:
1wget https://github.com/containerd/containerd/releases/download/v1.7.11/containerd-1.7.11-linux-amd64.tar.gz
将其解压缩到/usr/local下:
1tar Cxzvf /usr/local containerd-1.7.11-linux-amd64.tar.gz
2
3bin/
4bin/containerd-shim-runc-v2
5bin/ctr
6bin/containerd-shim
7bin/containerd-shim-runc-v1
8bin/containerd-stress
9bin/containerd
接下来从runc的github上单独下载安装runc，该二进制文件是静态构建的，并且应该适用于任何Linux发行版。
1wget https://github.com/opencontainers/runc/releases/download/v1.1.9/runc.amd64
2install -m 755 runc.amd64 /usr/local/sbin/runc
接下来生成containerd的配置文件:
1mkdir -p /etc/containerd
2containerd config default > /etc/containerd/config.toml
根据文档Container runtimes中的内容，对于使用systemd作为init system的Linux的发行版，使用systemd作为容器的cgroup driver可以确保服务器节点在资源紧张的情况更加稳定，因此这里配置各个节点上containerd的cgroup driver为systemd。
修改前面生成的配置文件/etc/containerd/config.toml：
1[plugins."io.containerd.grpc.v1.cri".containerd.runtimes.runc]
2 ...
3 [plugins."io.containerd.grpc.v1.cri".containerd.runtimes.runc.options]
4 SystemdCgroup = true
再修改/etc/containerd/config.toml中的
1[plugins."io.containerd.grpc.v1.cri"]
2 ...
3 # sandbox_image = "registry.k8s.io/pause:3.8"
4 sandbox_image = "registry.aliyuncs.com/google_containers/pause:3.9"
为了通过systemd启动containerd，请还需要从https://raw.githubusercontent.com/containerd/containerd/main/containerd.service下载containerd.service单元文件，并将其放置在 /etc/systemd/system/containerd.service中。
1cat << EOF > /etc/systemd/system/containerd.service
2[Unit]
3Description=containerd container runtime
4Documentation=https://containerd.io
5After=network.target local-fs.target
6
7[Service]
8ExecStartPre=-/sbin/modprobe overlay
9ExecStart=/usr/local/bin/containerd
10
11Type=notify
12Delegate=yes
13KillMode=process
14Restart=always
15RestartSec=5
16
17# Having non-zero Limit*s causes performance problems due to accounting overhead
18# in the kernel. We recommend using cgroups to do container-local accounting.
19LimitNPROC=infinity
20LimitCORE=infinity
21
22# Comment TasksMax if your systemd version does not supports it.
23# Only systemd 226 and above support this version.
24TasksMax=infinity
25OOMScoreAdjust=-999
26
27[Install]
28WantedBy=multi-user.target
29EOF
配置containerd开机启动，并启动containerd，执行以下命令:
1systemctl daemon-reload
2systemctl enable containerd --now
3systemctl status containerd
下载安装crictl工具:
1wget https://github.com/kubernetes-sigs/cri-tools/releases/download/v1.29.0/crictl-v1.29.0-linux-amd64.tar.gz
2tar -zxvf crictl-v1.29.0-linux-amd64.tar.gz
3install -m 755 crictl /usr/local/bin/crictl
使用crictl测试一下，确保可以打印出版本信息并且没有错误信息输出:
1crictl --runtime-endpoint=unix:///run/containerd/containerd.sock version
2
3Version: 0.1.0
4RuntimeName: containerd
5RuntimeVersion: v1.7.11
6RuntimeApiVersion: v1
2.使用kubeadm部署Kubernetes #
2.1 安装kubeadm和kubelet #
下面在各节点安装kubeadm和kubelet：
在Ubuntu系统上执行下面的命令:
1apt-get update
2apt-get install -y apt-transport-https ca-certificates curl gpg
3
4curl -fsSL https://pkgs.k8s.io/core:/stable:/v1.29/deb/Release.key | sudo gpg --dearmor -o /etc/apt/keyrings/kubernetes-apt-keyring.gpg
5
6
7echo 'deb [signed-by=/etc/apt/keyrings/kubernetes-apt-keyring.gpg] https://pkgs.k8s.io/core:/stable:/v1.29/deb/ /' | sudo tee /etc/apt/sources.list.d/kubernetes.list
8
9
10apt-get update
11
12apt install kubelet kubeadm kubectl
13
14apt-mark hold kubelet kubeadm kubectl
在openEuler和Rocky Linux系统中执行下面的命令:
1cat <<EOF | sudo tee /etc/yum.repos.d/kubernetes.repo
2[kubernetes]
3name=Kubernetes
4baseurl=https://pkgs.k8s.io/core:/stable:/v1.29/rpm/
5enabled=1
6gpgcheck=1
7gpgkey=https://pkgs.k8s.io/core:/stable:/v1.29/rpm/repodata/repomd.xml.key
8exclude=kubelet kubeadm kubectl cri-tools kubernetes-cni
9EOF
10
11
12yum makecache
13yum install -y kubelet kubeadm kubectl --disableexcludes=kubernetes
运行kubelet --help可以看到原来kubelet的绝大多数命令行flag参数都被DEPRECATED了，官方推荐我们使用--config指定配置文件，并在配置文件中指定原来这些flag所配置的内容。具体内容可以查看这里Set Kubelet parameters via a config file。最初Kubernetes这么做是为了支持动态Kubelet配置（Dynamic Kubelet Configuration），但动态Kubelet配置特性从k8s 1.22中已弃用，并在1.24中被移除。如果需要调整集群汇总所有节点kubelet的配置，还是推荐使用ansible等工具将配置分发到各个节点。
kubelet的配置文件必须是json或yaml格式，具体可查看这里。
在Kubernetes 1.22版本之前，Kubernetes没有为Linux系统提供NodeSwap功能，即默认要求关闭系统的Swap，如果不关闭，默认配置下kubelet将无法启动。
从Kubernetes 1.22开始引入了NodeSwap的Alpha支持，改功能在Kubernetes 1.28进入了Beta。
这里我们将激活NodeSwap功能，可以通过在启用NodeSwap feature gate来在节点上启用使用交换内存。此外，必须禁用failSwapOn配置设置，或者必须停用已弃用的--fail-swap-on命令行标志。
可以配置memorySwap.swapBehavior选项，以定义节点利用交换内存的方式。例如：
1# 此片段放入kubelet的配置文件
2memorySwap:
3 swapBehavior: UnlimitedSwap
swapBehavior的可用配置选项包括：
UnlimitedSwap（默认）：Kubernetes工作负载可以使用它们请求的所有交换内存，最多达到系统限制。
LimitedSwap：Kubernetes工作负载对交换内存的利用受到限制。只允许使用Burstable QoS(可突发QoS)的Pod使用Swap。
如果没有为memorySwap配置，并且启用了NodeSwap特性门，kubelet将默认应用与UnlimitedSwap设置相同的行为。
注意，NodeSwap仅支持cgroup v2。对于Kubernetes v1.28，不再支持与cgroup v1一起使用swap。
可以使用下面命令查看系统支持的cgroup版本:
1grep cgroup /proc/filesystems
2nodev cgroup
3nodev cgroup2
2.2 使用kubeadm init初始化集群 #
在各节点开机启动kubelet服务：
1systemctl enable kubelet.service
使用kubeadm config print init-defaults --component-configs KubeletConfiguration可以打印集群初始化默认的使用的配置：
1apiVersion: kubeadm.k8s.io/v1beta3
2bootstrapTokens:
3- groups:
4 - system:bootstrappers:kubeadm:default-node-token
5 token: abcdef.0123456789abcdef
6 ttl: 24h0m0s
7 usages:
8 - signing
9 - authentication
10kind: InitConfiguration
11localAPIEndpoint:
12 advertiseAddress: 1.2.3.4
13 bindPort: 6443
14nodeRegistration:
15 criSocket: unix:///var/run/containerd/containerd.sock
16 imagePullPolicy: IfNotPresent
17 name: node
18 taints: null
19---
20apiServer:
21 timeoutForControlPlane: 4m0s
22apiVersion: kubeadm.k8s.io/v1beta3
23certificatesDir: /etc/kubernetes/pki
24clusterName: kubernetes
25controllerManager: {}
26dns: {}
27etcd:
28 local:
29 dataDir: /var/lib/etcd
30imageRepository: registry.k8s.io
31kind: ClusterConfiguration
32kubernetesVersion: 1.29.0
33networking:
34 dnsDomain: cluster.local
35 serviceSubnet: 10.96.0.0/12
36scheduler: {}
37---
38apiVersion: kubelet.config.k8s.io/v1beta1
39authentication:
40 anonymous:
41 enabled: false
42 webhook:
43 cacheTTL: 0s
44 enabled: true
45 x509:
46 clientCAFile: /etc/kubernetes/pki/ca.crt
47authorization:
48 mode: Webhook
49 webhook:
50 cacheAuthorizedTTL: 0s
51 cacheUnauthorizedTTL: 0s
52cgroupDriver: systemd
53clusterDNS:
54- 10.96.0.10
55clusterDomain: cluster.local
56containerRuntimeEndpoint: ""
57cpuManagerReconcilePeriod: 0s
58evictionPressureTransitionPeriod: 0s
59fileCheckFrequency: 0s
60healthzBindAddress: 127.0.0.1
61healthzPort: 10248
62httpCheckFrequency: 0s
63imageMaximumGCAge: 0s
64imageMinimumGCAge: 0s
65kind: KubeletConfiguration
66logging:
67 flushFrequency: 0
68 options:
69 json:
70 infoBufferSize: "0"
71 verbosity: 0
72memorySwap: {}
73nodeStatusReportFrequency: 0s
74nodeStatusUpdateFrequency: 0s
75resolvConf: /run/systemd/resolve/resolv.conf
76rotateCertificates: true
77runtimeRequestTimeout: 0s
78shutdownGracePeriod: 0s
79shutdownGracePeriodCriticalPods: 0s
80staticPodPath: /etc/kubernetes/manifests
81streamingConnectionIdleTimeout: 0s
82syncFrequency: 0s
83volumeStatsAggPeriod: 0s
从默认的配置中可以看到，可以使用imageRepository定制在集群初始化时拉取k8s所需镜像的地址。基于默认配置定制出本次使用kubeadm初始化集群所需的配置文件kubeadm.yaml：
1apiVersion: kubeadm.k8s.io/v1beta3
2kind: InitConfiguration
3localAPIEndpoint:
4 advertiseAddress: 192.168.96.154
5 bindPort: 6443
6nodeRegistration:
7 criSocket: unix:///run/containerd/containerd.sock
8 taints:
9 - effect: PreferNoSchedule
10 key: node-role.kubernetes.io/master
11---
12apiVersion: kubeadm.k8s.io/v1beta3
13kind: ClusterConfiguration
14kubernetesVersion: 1.29.0
15imageRepository: registry.aliyuncs.com/google_containers
16networking:
17 podSubnet: 10.244.0.0/16
18---
19apiVersion: kubelet.config.k8s.io/v1beta1
20kind: KubeletConfiguration
21cgroupDriver: systemd
22failSwapOn: false
23---
24apiVersion: kubeproxy.config.k8s.io/v1alpha1
25kind: KubeProxyConfiguration
26mode: ipvs
这里定制了imageRepository为阿里云的registry，避免因gcr被墙，无法直接拉取镜像。criSocket设置了容器运行时为containerd。
同时设置kubelet的cgroupDriver为systemd，设置kube-proxy代理模式为ipvs。
在开始初始化集群之前可以使用kubeadm config images pull --config kubeadm.yaml预先在各个服务器节点上拉取所k8s需要的容器镜像。
1kubeadm config images list --config kubeadm.yaml
2registry.aliyuncs.com/google_containers/kube-apiserver:v1.29.0
3registry.aliyuncs.com/google_containers/kube-controller-manager:v1.29.0
4registry.aliyuncs.com/google_containers/kube-scheduler:v1.29.0
5registry.aliyuncs.com/google_containers/kube-proxy:v1.29.0
6registry.aliyuncs.com/google_containers/coredns:v1.11.1
7registry.aliyuncs.com/google_containers/pause:3.9
8registry.aliyuncs.com/google_containers/etcd:3.5.10-0
9
10kubeadm config images pull --config kubeadm.yaml
11[config/images] Pulled registry.aliyuncs.com/google_containers/kube-apiserver:v1.29.0
12[config/images] Pulled registry.aliyuncs.com/google_containers/kube-controller-manager:v1.29.0
13[config/images] Pulled registry.aliyuncs.com/google_containers/kube-scheduler:v1.29.0
14[config/images] Pulled registry.aliyuncs.com/google_containers/kube-proxy:v1.29.0
15[config/images] Pulled registry.aliyuncs.com/google_containers/coredns:v1.11.1
16[config/images] Pulled registry.aliyuncs.com/google_containers/pause:3.9
17[config/images] Pulled registry.aliyuncs.com/google_containers/etcd:3.5.10-0
接下来使用kubeadm初始化集群，选择node4作为Master Node，在node4上执行下面的命令：
1kubeadm init --config kubeadm.yaml
2
3[init] Using Kubernetes version: v1.29.0
4[preflight] Running pre-flight checks
5 [WARNING Swap]: swap is supported for cgroup v2 only; the NodeSwap feature gate of the kubelet is beta but disabled by default
6[preflight] Pulling images required for setting up a Kubernetes cluster
7[preflight] This might take a minute or two, depending on the speed of your internet connection
8[preflight] You can also perform this action in beforehand using 'kubeadm config images pull'
9[certs] Using certificateDir folder "/etc/kubernetes/pki"
10[certs] Using existing ca certificate authority
11[certs] Using existing apiserver certificate and key on disk
12[certs] Using existing apiserver-kubelet-client certificate and key on disk
13[certs] Using existing front-proxy-ca certificate authority
14[certs] Using existing front-proxy-client certificate and key on disk
15[certs] Using existing etcd/ca certificate authority
16[certs] Using existing etcd/server certificate and key on disk
17[certs] Using existing etcd/peer certificate and key on disk
18[certs] Using existing etcd/healthcheck-client certificate and key on disk
19[certs] Using existing apiserver-etcd-client certificate and key on disk
20[certs] Using the existing "sa" key
21[kubeconfig] Using kubeconfig folder "/etc/kubernetes"
22[kubeconfig] Writing "admin.conf" kubeconfig file
23[kubeconfig] Writing "super-admin.conf" kubeconfig file
24[kubeconfig] Writing "kubelet.conf" kubeconfig file
25[kubeconfig] Writing "controller-manager.conf" kubeconfig file
26[kubeconfig] Writing "scheduler.conf" kubeconfig file
27[etcd] Creating static Pod manifest for local etcd in "/etc/kubernetes/manifests"
28[control-plane] Using manifest folder "/etc/kubernetes/manifests"
29[control-plane] Creating static Pod manifest for "kube-apiserver"
30[control-plane] Creating static Pod manifest for "kube-controller-manager"
31[control-plane] Creating static Pod manifest for "kube-scheduler"
32[kubelet-start] Writing kubelet environment file with flags to file "/var/lib/kubelet/kubeadm-flags.env"
33[kubelet-start] Writing kubelet configuration to file "/var/lib/kubelet/config.yaml"
34[kubelet-start] Starting the kubelet
35[wait-control-plane] Waiting for the kubelet to boot up the control plane as static Pods from directory "/etc/kubernetes/manifests". This can take up to 4m0s
36[apiclient] All control plane components are healthy after 12.006365 seconds
37[upload-config] Storing the configuration used in ConfigMap "kubeadm-config" in the "kube-system" Namespace
38[kubelet] Creating a ConfigMap "kubelet-config" in namespace kube-system with the configuration for the kubelets in the cluster
39[upload-certs] Skipping phase. Please see --upload-certs
40[mark-control-plane] Marking the node node4 as control-plane by adding the labels: [node-role.kubernetes.io/control-plane node.kubernetes.io/exclude-from-external-load-balancers]
41[mark-control-plane] Marking the node node4 as control-plane by adding the taints [node-role.kubernetes.io/master:PreferNoSchedule]
42[bootstrap-token] Using token: alhelp.oyjw8wk6zyw5b55p
43[bootstrap-token] Configuring bootstrap tokens, cluster-info ConfigMap, RBAC Roles
44[bootstrap-token] Configured RBAC rules to allow Node Bootstrap tokens to get nodes
45[bootstrap-token] Configured RBAC rules to allow Node Bootstrap tokens to post CSRs in order for nodes to get long term certificate credentials
46[bootstrap-token] Configured RBAC rules to allow the csrapprover controller automatically approve CSRs from a Node Bootstrap Token
47[bootstrap-token] Configured RBAC rules to allow certificate rotation for all node client certificates in the cluster
48[bootstrap-token] Creating the "cluster-info" ConfigMap in the "kube-public" namespace
49[kubelet-finalize] Updating "/etc/kubernetes/kubelet.conf" to point to a rotatable kubelet client certificate and key
50[addons] Applied essential addon: CoreDNS
51[addons] Applied essential addon: kube-proxy
52
53Your Kubernetes control-plane has initialized successfully!
54
55To start using your cluster, you need to run the following as a regular user:
56
57 mkdir -p $HOME/.kube
58 sudo cp -i /etc/kubernetes/admin.conf $HOME/.kube/config
59 sudo chown $(id -u):$(id -g) $HOME/.kube/config
60
61Alternatively, if you are the root user, you can run:
62
63 export KUBECONFIG=/etc/kubernetes/admin.conf
64
65You should now deploy a pod network to the cluster.
66Run "kubectl apply -f [podnetwork].yaml" with one of the options listed at:
67 https://kubernetes.io/docs/concepts/cluster-administration/addons/
68
69Then you can join any number of worker nodes by running the following on each as root:
70
71kubeadm join 192.168.96.154:6443 --token alhelp.oyjw8wk6zyw5b55p \
72 --discovery-token-ca-cert-hash sha256:402b5d2d29367ada9b8ee2b37bfb246a318cbfce71d9e38c9117701455714f3e
上面记录了完成的初始化输出的内容，根据输出的内容基本上可以看出手动初始化安装一个Kubernetes集群所需要的关键步骤。 其中有以下关键内容：
[certs]生成相关的各种证书[kubeconfig]生成相关的kubeconfig文件[kubelet-start]生成kubelet的配置文件"/var/lib/kubelet/config.yaml"[control-plane]使用/etc/kubernetes/manifests目录中的yaml文件创建apiserver、controller-manager、scheduler的静态pod[bootstraptoken]生成token记录下来，后边使用kubeadm join往集群中添加节点时会用到[addons]安装基本插件:CoreDNS,kube-proxy- 下面的命令是配置常规用户如何使用kubectl访问集群：
1mkdir -p $HOME/.kube 2sudo cp -i /etc/kubernetes/admin.conf $HOME/.kube/config 3sudo chown $(id -u):$(id -g) $HOME/.kube/config - 最后给出了将另外2个节点加入集群的命令:
1kubeadm join 192.168.96.154:6443 --token alhelp.oyjw8wk6zyw5b55p \
2 --discovery-token-ca-cert-hash sha256:402b5d2d29367ada9b8ee2b37bfb246a318cbfce71d9e38c9117701455714f3e
查看一下集群状态，确认个组件都处于healthy状态，结果出现了错误:
1kubectl get cs
2
3Warning: v1 ComponentStatus is deprecated in v1.19+
4NAME STATUS MESSAGE ERROR
5controller-manager Healthy ok
6scheduler Healthy ok
7etcd-0 Healthy ok
8
9
10kubectl get --raw='/readyz?verbose'
11...
12readyz check passed
13
14kubectl get --raw='/livez?verbose'
15...
16livez check passed
集群初始化如果遇到问题，可以使用kubeadm reset命令进行清理。
2.3 安装包管理器helm 3 #
Helm是Kubernetes的包管理器，后续流程也将使用Helm安装Kubernetes的常用组件。 这里先在master节点node4上安装helm。
1wget https://get.helm.sh/helm-v3.13.3-linux-amd64.tar.gz
2tar -zxvf helm-v3.13.3-linux-amd64.tar.gz
3install -m 755 linux-amd64/helm /usr/local/bin/helm
执行helm list确认没有错误输出。
2.4 部署Pod Network组件Calico #
选择calico作为k8s的Pod网络组件，下面使用helm在k8s集群中安装calico。
下载tigera-operator的helm chart:
1wget https://github.com/projectcalico/calico/releases/download/v3.27.0/tigera-operator-v3.27.0.tgz
查看这个chart的中可定制的配置:
1helm show values tigera-operator-v3.27.0.tgz
2
3# imagePullSecrets is a special helm field which, when specified, creates a secret
4# containing the pull secret which is used to pull all images deployed by this helm chart and the resulting operator.
5# this field is a map where the key is the desired secret name and the value is the contents of the imagePullSecret.
6#
7# Example: --set-file imagePullSecrets.gcr=./pull-secret.json
8imagePullSecrets: {}
9
10installation:
11 enabled: true
12 kubernetesProvider: ""
13 # imagePullSecrets are configured on all images deployed by the tigera-operator.
14 # secrets specified here must exist in the tigera-operator namespace; they won't be created by the operator or helm.
15 # imagePullSecrets are a slice of LocalObjectReferences, which is the same format they appear as on deployments.
16 #
17 # Example: --set installation.imagePullSecrets[0].name=my-existing-secret
18 imagePullSecrets: []
19
20apiServer:
21 enabled: true
22
23certs:
24 node:
25 key:
26 cert:
27 commonName:
28 typha:
29 key:
30 cert:
31 commonName:
32 caBundle:
33
34# Resource requests and limits for the tigera/operator pod.
35resources: {}
36
37# Tolerations for the tigera/operator pod.
38tolerations:
39- effect: NoExecute
40 operator: Exists
41- effect: NoSchedule
42 operator: Exists
43
44# NodeSelector for the tigera/operator pod.
45nodeSelector:
46 kubernetes.io/os: linux
47
48# Custom annotations for the tigera/operator pod.
49podAnnotations: {}
50
51# Custom labels for the tigera/operator pod.
52podLabels: {}
53
54# Image and registry configuration for the tigera/operator pod.
55tigeraOperator:
56 image: tigera/operator
57 version: v1.32.3
58 registry: quay.io
59calicoctl:
60 image: docker.io/calico/ctl
61 tag: v3.27.0
62
63kubeletVolumePluginPath: /var/lib/kubelet
定制的values.yaml如下:
1# 可针对上面的配置进行定制,例如calico的镜像改成从私有库拉取。
2# 这里只是个人本地环境测试k8s新版本，因此只有下面几行配置
3apiServer:
4 enabled: false
5installation:
6 kubeletVolumePluginPath: None
使用helm安装calico：
1helm install calico tigera-operator-v3.27.0.tgz -n kube-system --create-namespace -f values.yaml
等待并确认所有pod处于Running状态:
1kubectl get pod -n kube-system | grep tigera-operator
2tigera-operator-55585899bf-qkr84 1/1 Running 0 26s
3
4kubectl get pods -n calico-system
5NAME READY STATUS RESTARTS AGE
6calico-kube-controllers-6784546df7-5dzld 1/1 Running 0 5m30s
7calico-node-24px9 1/1 Running 0 5m30s
8calico-typha-75854bc9c9-5zvrb 1/1 Running 0 5m31s
9csi-node-driver-mttxs 2/2 Running 0 5m30s
查看一下calico向k8s中添加的api资源:
1kubectl api-resources | grep calico
2bgpconfigurations crd.projectcalico.org/v1 false BGPConfiguration
3bgpfilters crd.projectcalico.org/v1 false BGPFilter
4bgppeers crd.projectcalico.org/v1 false BGPPeer
5blockaffinities crd.projectcalico.org/v1 false BlockAffinity
6caliconodestatuses crd.projectcalico.org/v1 false CalicoNodeStatus
7clusterinformations crd.projectcalico.org/v1 false ClusterInformation
8felixconfigurations crd.projectcalico.org/v1 false FelixConfiguration
9globalnetworkpolicies crd.projectcalico.org/v1 false GlobalNetworkPolicy
10globalnetworksets crd.projectcalico.org/v1 false GlobalNetworkSet
11hostendpoints crd.projectcalico.org/v1 false HostEndpoint
12ipamblocks crd.projectcalico.org/v1 false IPAMBlock
13ipamconfigs crd.projectcalico.org/v1 false IPAMConfig
14ipamhandles crd.projectcalico.org/v1 false IPAMHandle
15ippools crd.projectcalico.org/v1 false IPPool
16ipreservations crd.projectcalico.org/v1 false IPReservation
17kubecontrollersconfigurations crd.projectcalico.org/v1 false KubeControllersConfiguration
18networkpolicies crd.projectcalico.org/v1 true NetworkPolicy
19networksets crd.projectcalico.org/v1 true NetworkSet
这些api资源是属于calico的，因此不建议使用kubectl来管理，推荐按照calicoctl来管理这些api资源。 将calicoctl安装为kubectl的插件:
1cd /usr/local/bin
2curl -o kubectl-calico -O -L "https://github.com/projectcalico/calico/releases/download/v3.27.0/calicoctl-linux-amd64"
3chmod +x kubectl-calico
验证插件正常工作:
1kubectl calico -h
2.5 验证k8s DNS是否可用 #
1kubectl run curl --image=radial/busyboxplus:curl -it
2If you don't see a command prompt, try pressing enter.
3[ root@curl:/ ]$
进入后执行nslookup kubernetes.default确认解析正常:
1nslookup kubernetes.default
2Server: 10.96.0.10
3Address 1: 10.96.0.10 kube-dns.kube-system.svc.cluster.local
4
5Name: kubernetes.default
6Address 1: 10.96.0.1 kubernetes.default.svc.cluster.local
2.6 向Kubernetes集群中添加Node节点 #
下面将node5, node6添加到Kubernetes集群中，分别在node5, node6上执行:
1kubeadm join 192.168.96.154:6443 --token alhelp.oyjw8wk6zyw5b55p \
2 --discovery-token-ca-cert-hash sha256:402b5d2d29367ada9b8ee2b37bfb246a318cbfce71d9e38c9117701455714f3e
node5和node6加入集群时遇到了如下问题，调度到node5(openEuler 22.03系统)或node6(Rock Linux 8.8)的calico-typha Pod无法启动，并报下面的错误:
1kubectl describe po calico-node-ht7cf -n calico-system
2...
3kubelet Failed to create pod sandbox: open /run/systemd/resolve/resolv.conf: no such file or directory
而/run/systemd/resolve/resolv.conf 文件是由systemd-resolved服务管理的。Ubuntu 22.04上默认安装并启动了这个服务。openEuler 22.03上没有安装这个服务。Rocky Linux 8.8上默认安装但没有启动这个服务。
下面node5上安装并启动systemd-resolved:
1yum install -y systemd-resolved
2systemctl enable systemd-resolved --now
3systemctl status systemd-resolved
在node6上启动systemd-resolved:
1systemctl enable systemd-resolved --now
2systemctl status systemd-resolved
之后3个节点上的calico相关pod全部启动正常:
1kubectl get po -n calico-system -o wide
2NAME READY STATUS RESTARTS AGE IP NODE NOMINATED NODE READINESS GATES
3calico-kube-controllers-6784546df7-5dzld 1/1 Running 0 20m 10.244.3.67 node4 <none> <none>
4calico-node-24px9 1/1 Running 0 20m 192.168.96.154 node4 <none> <none>
5calico-node-ht7cf 1/1 Running 0 7m6s 192.168.96.155 node5 <none> <none>
6calico-node-tzql5 1/1 Running 0 7m4s 192.168.96.156 node6 <none> <none>
7calico-typha-75854bc9c9-5zvrb 1/1 Running 0 20m 192.168.96.154 node4 <none> <none>
8calico-typha-75854bc9c9-l8bmh 1/1 Running 0 6m56s 192.168.96.155 node5 <none> <none>
9csi-node-driver-2mk5s 2/2 Running 0 7m6s 10.244.33.129 node5 <none> <none>
10csi-node-driver-9qrzx 2/2 Running 0 7m4s 10.244.139.1 node6 <none> <none>
11csi-node-driver-mttxs 2/2 Running 0 20m 10.244.3.65 node4 <none> <none>
在master节点上执行命令查看集群中的节点(需要等待新加入节点上的calico-node pod启动正常)：
1kubectl get node
2NAME STATUS ROLES AGE VERSION
3node4 Ready control-plane 34m v1.29.0
4node5 Ready <none> 7m31s v1.29.0
5node6 Ready <none> 7m29s v1.29.0
3.Kubernetes常用组件部署 #
3.1 使用Helm部署ingress-nginx #
为了便于将集群中的服务暴露到集群外部，需要使用Ingress。接下来使用Helm将ingress-nginx部署到Kubernetes上。 Nginx Ingress Controller被部署在Kubernetes的边缘节点上。
这里将node4(192.168.96.154)作为边缘节点，打上Label：
1kubectl label node node4 node-role.kubernetes.io/edge=
下载ingress-nginx的helm chart:
1wget https://github.com/kubernetes/ingress-nginx/releases/download/helm-chart-4.9.0/ingress-nginx-4.9.0.tgz
查看ingress-nginx-4.9.0.tgz这个chart的可定制配置:
1helm show values ingress-nginx-4.9.0.tgz
对values.yaml配置定制如下:
1controller:
2 ingressClassResource:
3 name: nginx
4 enabled: true
5 default: true
6 controllerValue: "k8s.io/ingress-nginx"
7 admissionWebhooks:
8 enabled: false
9 replicaCount: 1
10 image:
11 # registry: registry.k8s.io
12 # image: ingress-nginx/controller
13 # tag: "v1.9.5"
14 registry: docker.io
15 image: unreachableg/registry.k8s.io_ingress-nginx_controller
16 tag: "v1.9.5"
17 digest: sha256:bdc54c3e73dcec374857456559ae5757e8920174483882b9e8ff1a9052f96a35
18 hostNetwork: true
19 nodeSelector:
20 node-role.kubernetes.io/edge: ''
21 affinity:
22 podAntiAffinity:
23 requiredDuringSchedulingIgnoredDuringExecution:
24 - labelSelector:
25 matchExpressions:
26 - key: app
27 operator: In
28 values:
29 - nginx-ingress
30 - key: component
31 operator: In
32 values:
33 - controller
34 topologyKey: kubernetes.io/hostname
35 tolerations:
36 - key: node-role.kubernetes.io/master
37 operator: Exists
38 effect: NoSchedule
39 - key: node-role.kubernetes.io/master
40 operator: Exists
41 effect: PreferNoSchedule
nginx ingress controller的副本数replicaCount为1，将被调度到node4这个边缘节点上。这里并没有指定nginx ingress controller service的externalIPs，而是通过hostNetwork: true设置nginx ingress controller使用宿主机网络。
因为registry.k8s.io被墙，这里替换成unreachableg/registry.k8s.io_ingress-nginx_controller提前拉取一下镜像:
1crictl --runtime-endpoint=unix:///run/containerd/containerd.sock pull unreachableg/registry.k8s.io_ingress-nginx_controller:v1.9.5
1helm install ingress-nginx ingress-nginx-4.9.0.tgz --create-namespace -n ingress-nginx -f values.yaml
1kubectl get po -n ingress-nginx
2NAME READY STATUS RESTARTS AGE
3ingress-nginx-controller-6445445cb8-c4fh4 1/1 Running 0 74s
测试访问http://192.168.96.154返回默认的nginx 404页，则部署完成。
3.2 使用Helm部署dashboard #
先部署metrics-server：
1wget https://github.com/kubernetes-sigs/metrics-server/releases/download/v0.6.4/components.yaml
修改components.yaml中的image为docker.io/unreachableg/k8s.gcr.io_metrics-server_metrics-server:v0.6.4。
修改components.yaml中容器的启动参数，加入--kubelet-insecure-tls。
1kubectl apply -f components.yaml
metrics-server的pod正常启动后，等一段时间就可以使用kubectl top查看集群和pod的metrics信息:
1kubectl top node
2NAME CPU(cores) CPU% MEMORY(bytes) MEMORY%
3node4 373m 9% 2184Mi 27%
4node5 42m 1% 968Mi 12%
5node6 131m 3% 918Mi 12%
6
7kubectl top pod -n kube-system
8NAME CPU(cores) MEMORY(bytes)
9coredns-857d9ff4c9-5pvft 6m 13Mi
10coredns-857d9ff4c9-zkmm6 5m 13Mi
11etcd-node4 62m 58Mi
12kube-apiserver-node4 154m 350Mi
13kube-controller-manager-node4 35m 45Mi
14kube-proxy-4qfvt 31m 19Mi
15kube-proxy-98k25 9m 17Mi
16kube-proxy-rbh22 9m 18Mi
17kube-scheduler-node4 9m 16Mi
18metrics-server-7d686f4d9d-pxn8g 13m 18Mi
19tigera-operator-55585899bf-qkr84 5m 28Mi
接下来使用helm部署k8s的dashboard。当前k8s dashboard已经更新到了v3.0.0-alpha0，这里体验一下v3版本。
从k8s dashboard的v3版本开始，底层架构已更改，需要进行干净的安装，如果是在做升级dashboard操作，请首先移除先前的安装，这里是全新安装可以忽略。
k8s dashboard的v3版本现在默认使用cert-manager和nginx-ingress-controller。如果选择基于yaml清单的安装，请确保在集群中已安装它们。
我们前面已经安装了nginx-ingress-controller，下面先安装cert-manager:
1wget https://github.com/cert-manager/cert-manager/releases/download/v1.13.3/cert-manager.yaml
2
3kubectl apply -f cert-manager.yaml
确保cert-manager的所有pod启动正常:
1kubectl get po -n cert-manager
2NAME READY STATUS RESTARTS AGE
3cert-manager-6774cd657f-q9qpf 1/1 Running 0 102s
4cert-manager-cainjector-55c8b7b49b-vf8r4 1/1 Running 0 102s
5cert-manager-webhook-57797c469d-cgw4n 1/1 Running 0 102s
下载dashboard的yaml清单文件:
1wget https://raw.githubusercontent.com/kubernetes/dashboard/v3.0.0-alpha0/charts/kubernetes-dashboard.yaml
编辑kubernetes-dashboard.yaml清单文件，将其中的ingress中的host替换想分配给k8s dashboard的域名:
1kind: Ingress
2apiVersion: networking.k8s.io/v1
3metadata:
4 name: kubernetes-dashboard
5 namespace: kubernetes-dashboard
6 labels:
7 app.kubernetes.io/name: nginx-ingress
8 app.kubernetes.io/part-of: kubernetes-dashboard
9 annotations:
10 nginx.ingress.kubernetes.io/ssl-redirect: "true"
11 cert-manager.io/issuer: selfsigned
12spec:
13 ingressClassName: nginx
14 tls:
15 - hosts:
16 - localhost
17 secretName: kubernetes-dashboard-certs
18 rules:
19 - host: k8s.example.com
20 http:
21 paths:
22 - path: /
23 pathType: Prefix
24 backend:
25 service:
26 name: kubernetes-dashboard-web
27 port:
28 name: web
29 - path: /api
30 pathType: Prefix
31 backend:
32 service:
33 name: kubernetes-dashboard-api
34 port:
35 name: api
这里将k8s以ingress暴露到k8s集群外边，这里模拟真实环境中以域名形式访问暴露的服务。k8s dashboard的ingress中配置的域名为k8s.example.com，实际上需要客户端浏览器电脑的DNS可以解析到ingress controller(这里是192.168.96.154)上。
如果没有DNS，可以手动在客户端浏览器电脑设置hosts配置192.168.96.154 k8s.example.com。
安装dashboard的yaml清单文件:
1kubectl apply -f kubernetes-dashboard.yaml
确认dashboard的相关Pod启动正常:
1kubectl get po -n kubernetes-dashboard
2NAME READY STATUS RESTARTS AGE
3kubernetes-dashboard-api-8586787f7-txzdx 1/1 Running 0 3m40s
4kubernetes-dashboard-metrics-scraper-6959b784dc-424p5 1/1 Running 0 3m40s
5kubernetes-dashboard-web-6b6d549b4-jcp2l 1/1 Running 0 3m40s
6
7kubectl get ingress -n kubernetes-dashboard
8NAME CLASS HOSTS ADDRESS PORTS AGE
9kubernetes-dashboard nginx k8s.example.com 80, 443 3m49s
创建管理员sa:
1kubectl create serviceaccount kube-dashboard-admin-sa -n kube-system
2
3kubectl create clusterrolebinding kube-dashboard-admin-sa \
4--clusterrole=cluster-admin --serviceaccount=kube-system:kube-dashboard-admin-sa
创建集群管理员登录dashboard所需token:
1kubectl create token kube-dashboard-admin-sa -n kube-system --duration=87600h
2
3eyJhbGciOiJSUzI1NiIsImtpZCI6Im5SWVpMcGZMcHFjYVdFcFNzX2kwTmwxYUx2M2NRckU5MFJBUmpSLW1fV28ifQ.eyJhdWQiOlsiaHR0cHM6Ly9rdWJlcm5ldGVzLmRlZmF1bHQuc3ZjLmNsdXN0ZXIubG9jYWwiXSwiZXhwIjoyMDE5MTExNjI5LCJpYXQiOjE3MDM3NTE2MjksImlzcyI6Imh0dHBzOi8va3ViZXJuZXRlcy5kZWZhdWx0LnN2Yy5jbHVzdGVyLmxvY2FsIiwia3ViZXJuZXRlcy5pbyI6eyJuYW1lc3BhY2UiOiJrdWJlLXN5c3RlbSIsInNlcnZpY2VhY2NvdW50Ijp7Im5hbWUiOiJrdWJlLWRhc2hib2FyZC1hZG1pbi1zYSIsInVpZCI6ImQ1YzZiMDdmLWUzMDAtNDMzOS04ZDY1LTUwYzg0N2FjMjg2MCJ9fSwibmJmIjoxNzAzNzUxNjI5LCJzdWIiOiJzeXN0ZW06c2VydmljZWFjY291bnQ6a3ViZS1zeXN0ZW06a3ViZS1kYXNoYm9hcmQtYWRtaW4tc2EifQ.EOFDNd0GvXjJpoUYFjOKDhuEbSJgLn6RuQeBgwjN-C4lR5C0URwXVarDUmGJTJZiAcHsajM1RGmR9u26vFvh9ZKTaQOkpJKYvJACiUwiOFZzGv_j2Cc5erZbiJskNMzl_Yt_fyACDpZpB20pjtT5e91C5Z7NPdgHbQsKt0Nkj6iLoIrGDihWBUEl33v1q1JixYyvtr9v2TcmmT8kQDmwluIsetW2TwN17ZVD1wsVz9iRgu0xwEWgzKh9FebQKJOsMmKWerca9ov_PD62ppElR0553-spgjjxow-rZ4mxn3u5M-dPfX57yIBQjczCd3jyEDedMs_RmRxUz_rtebdQAw
使用上面的token登录k8s dashboard。
