kubeadm系列-01-preflight究竟有多少check

我们知道 kubeadm init 的过程中会进行很多 preflight 的检查，这些主要是指内核参数、模块、CRI 等环境的检查，如果有哪些配置不符合 Kubernetes 的要求，就会抛出 Warning 或者 Error 的信息，下面就是 preflight 的主要逻辑

// Checker validates the state of the system to ensure kubeadm will be
// successful as often as possible.
type Checker interface {
	Check() (warnings, errorList []error)
	Name() string
}

复制

如果有 diy 的 check 需求，可以在代码里继承这个接口进行扩展，下面举个 check 的例子，很明显 ContainerRuntimeCheck 是对 CRI 也就是容器运行时进行的检查

// ContainerRuntimeCheck verifies the container runtime.
type ContainerRuntimeCheck struct {
	runtime utilruntime.ContainerRuntime
}

// Name returns label for RuntimeCheck.
func (ContainerRuntimeCheck) Name() string {
	return "CRI"
}

// Check validates the container runtime
func (crc ContainerRuntimeCheck) Check() (warnings, errorList []error) {
	klog.V(1).Infoln("validating the container runtime")
	if err := crc.runtime.IsRunning(); err != nil {
		errorList = append(errorList, err)
	}
	return warnings, errorList
}

复制

而真正起到检查的作用是下面这个函数，其实就是宿主机执行一下 crictl info，并且接收其返回，老铁们不妨在宿主机上直接运行一下看看结果

// IsRunning checks if runtime is running
func (runtime *CRIRuntime) IsRunning() error {
	if out, err := runtime.crictl("info").CombinedOutput(); err != nil {
		return errors.Wrapf(err, "container runtime is not running: output: %s, error", string(out))
	}
	return nil
}

复制

所有的 Check 里面是会有小部分交错的部分，比如说检查防火墙的问题，先会对 Firewall 这个服务做 service check，然后才会对具体的端口进行检查

下面是所有 check 的统计

1. CRI: 检查容器运行时是否有在运行
2. Service: 检查是否enable和active
3. Firewall: 检查防火墙是否有关闭
4. Port: 检查某些端口是否有放开
5. Privileged: 检查一些权限的问题
6. Dir Available: 检查目录是否有效
7. File Available: 检查文件是否有效
8. File Existing: 检查文件是否存在
9. File Content: 检查文件中是否有指定的内容
10. In Path: 检查某些可执行文件是否在指定的目录
11. Hostname: 检查主机名的格式
12. HTTP Proxy: 检查本机是否有Proxy设置
13. HTTP Proxy CIDR: 检查本机有哪些地址会走Proxy
14. System Verification: 检查系统版本
15. Kubernetes Version: 检查Kubernetes的版本
16. Kubelet Version: 检查Kubelet的版本
17. SwapCheck: 检查Swap是否关闭
18. External Etcd Version: 检查外部etcd的版本
19. Image Pull: 检查镜像仓库是否连通
20. Num CPU: 检查本机CPU数量是否符合kubeadm的最低要求
21. Mem: 检查本机内存是否符合kubeadm的最低要求

真正在做检查的时候，还会区分是 controlplane 还是普通的 worker 节点，不同角色具体要做的检查是不尽相同的

我们看一下 In Path 这个检查，也就是检查一些必要的二进制文件或者命令是否已经安装，另外还要看 mandatory 如果是 true 的话，那就是必须要符合的，否则就是可有可无，不过如果没有就会提示出来，会建议用户去安装的

InPathCheck{executable: "crictl", mandatory: true, exec: execer},
InPathCheck{executable: "conntrack", mandatory: true, exec: execer},
InPathCheck{executable: "ip", mandatory: true, exec: execer},
InPathCheck{executable: "iptables", mandatory: true, exec: execer},
InPathCheck{executable: "mount", mandatory: true, exec: execer},
InPathCheck{executable: "nsenter", mandatory: true, exec: execer},
InPathCheck{executable: "ebtables", mandatory: false, exec: execer},
InPathCheck{executable: "ethtool", mandatory: false, exec: execer},
InPathCheck{executable: "socat", mandatory: false, exec: execer},
InPathCheck{executable: "tc", mandatory: false, exec: execer},
InPathCheck{executable: "touch", mandatory: false, exec: execer})

复制

最后我们看一下 System Verification，主要是针对主机的系统来进行一些模块的检测，我们主要看一下 Linux 下的检查，内核很多模块有以及没有，还是有较大的差别的，所以不要轻视这部分的检查，以为主要是 Linux 系统就没啥问题了，有时候恰恰是这部分的内容更难排查

// DefaultSysSpec is the default SysSpec for Linux
var DefaultSysSpec = SysSpec{
	OS: "Linux",
	KernelSpec: KernelSpec{
		Versions: []string{`^3\.[1-9][0-9].*$`, `^([4-9]|[1-9][0-9]+)\.([0-9]+)\.([0-9]+).*$`}, // Requires 3.10+, or newer
		// TODO(random-liu): Add more config
		// TODO(random-liu): Add description for each kernel configuration:
		Required: []KernelConfig{
			{Name: "NAMESPACES"},
			{Name: "NET_NS"},
			{Name: "PID_NS"},
			{Name: "IPC_NS"},
			{Name: "UTS_NS"},
			{Name: "CGROUPS"},
			{Name: "CGROUP_CPUACCT"},
			{Name: "CGROUP_DEVICE"},
			{Name: "CGROUP_FREEZER"},
			{Name: "CGROUP_PIDS"},
			{Name: "CGROUP_SCHED"},
			{Name: "CPUSETS"},
			{Name: "MEMCG"},
			{Name: "INET"},
			{Name: "EXT4_FS"},
			{Name: "PROC_FS"},
			{Name: "NETFILTER_XT_TARGET_REDIRECT", Aliases: []string{"IP_NF_TARGET_REDIRECT"}},
			{Name: "NETFILTER_XT_MATCH_COMMENT"},
			{Name: "FAIR_GROUP_SCHED"},
		},
		Optional: []KernelConfig{
			{Name: "OVERLAY_FS", Aliases: []string{"OVERLAYFS_FS"}, Description: "Required for overlayfs."},
			{Name: "AUFS_FS", Description: "Required for aufs."},
			{Name: "BLK_DEV_DM", Description: "Required for devicemapper."},
			{Name: "CFS_BANDWIDTH", Description: "Required for CPU quota."},
			{Name: "CGROUP_HUGETLB", Description: "Required for hugetlb cgroup."},
			{Name: "SECCOMP", Description: "Required for seccomp."},
			{Name: "SECCOMP_FILTER", Description: "Required for seccomp mode 2."},
		},
		Forbidden: []KernelConfig{},
	},
	Cgroups: []string{"cpu", "cpuacct", "cpuset", "devices", "freezer", "memory", "pids"},
	CgroupsOptional: []string{
		// The hugetlb cgroup is optional since some kernels are compiled without support for huge pages
		// and therefore lacks corresponding hugetlb cgroup
		"hugetlb",
		// The blkio cgroup is optional since some kernels are compiled without support for block I/O throttling.
		// Containerd and cri-o will use blkio to track disk I/O and throttling in both cgroup v1 and v2.
		"blkio",
	},
	CgroupsV2:         []string{"cpu", "cpuset", "devices", "freezer", "memory", "pids"},
	CgroupsV2Optional: []string{"hugetlb", "blkio"},
	RuntimeSpec: RuntimeSpec{
		DockerSpec: &DockerSpec{
			Version:     []string{`1\.1[1-3]\..*`, `17\.0[3,6,9]\..*`, `18\.0[6,9]\..*`, `19\.03\..*`, `20\.10\..*`},
			GraphDriver: []string{"aufs", "btrfs", "overlay", "overlay2", "devicemapper", "zfs"},
		},
	},
}

复制