Through vim.vm.CloneSpec() cloning failed

[snow-with-tea]

Describe the bug

I have a virtual machine named "template_centos7", and now I want to clone it, but I have failed using the following script

Did I make a mistake in writing somewhere

import logging
import re
import sys
import random
import time
from pyVim.connect import SmartConnect, Disconnect
from pyVmomi import vim
import ssl

class exsi:
    def __init__(self, host="127.0.0.1", user="root", pwd="12345678", ):
        # 禁用 SSL 证书验证（不推荐在生产环境中使用）
        sslContext = ssl.create_default_context()
        sslContext.check_hostname = False
        sslContext.verify_mode = ssl.CERT_NONE

        # 连接到 exsi
        self.si = SmartConnect(host=host, user=user, pwd=pwd, sslContext=sslContext)
        # 检索内容
        self.content = self.si.RetrieveContent()
        for child in self.content.rootFolder.childEntity:
            print(child.name)
            vm_folder = child.vmFolder  # child is a datacenter
        # rootFolder 是 vCenter Server 中所有对象的根目录，包括虚拟机、主机、数据中心等。我们将这个根目录赋值给 container，以便稍后在容器视图中使用
        container = self.content.rootFolder
        # vim.VirtualMachine 是 PyVmomi 中表示虚拟机的类
        # 创建了一个列表 view_type, 其中包含我们感兴趣的对象类型, 这里只填写了虚拟机类型
        view_type = [vim.VirtualMachine]
        # recursive 是一个布尔值, 表示是否递归地检索容器中的对象
        recursive = True
        # 创建容器视图, 三个参数分别为:
        #     container：容器视图的根节点
        #     view_type：感兴趣的对象类型, 即虚拟机。
        #     recursive：是否递归地检索对象。
        self.container_view = self.content.viewManager.CreateContainerView(container, view_type, recursive)

    def get_attr(self, obj, obj_attrs: str, default=None):
        """
        getattr不知为何不能获取vm的多层属性, 此函数模仿getattr获取类的多层属性, 以小数点为分隔
        本函数不支持获取不到属性时报错退出, 只能返回一个默认值, 这个值默认为空
        :param obj: 要解析的对象
        :param obj_attrs: 对象的属性, 多层之间以小数点分隔
        :param default: 没有找到属性时的默认值, 可以不填, 不填为空
        """
        list_attrs = re.split('\\.', obj_attrs)
        obj_init = obj
        for i in range(0, len(list_attrs)):
            obj_init = getattr(obj_init, list_attrs[i], None)
            # print(obj_init)
            if not obj_init:
                obj_init = default
                break
        return obj_init

    def generate_mac(self):
        """
        生成新的mac地址
        """
        random.seed(time.time())
        # 生成MAC地址的前三个字节（厂商标识）
        oui = [random.randint(0x00, 0x7f) for _ in range(3)]
        # 生成MAC地址的后三个字节（设备标识）
        nic = [random.randint(0x00, 0xff) for _ in range(3)]
        # 将字节转换为十六进制字符串并连接起来
        mac_address = ':'.join(['{:02x}'.format(b) for b in oui + nic]).upper()
        return mac_address

    def wait_for_task(self, task):
        """
        等待exsi任务执行结束并获取结果
        """
        task_done = False
        # 死循环检测task的状态
        while not task_done:
            # logging.info(f"task.....{task.info.state} ")
            if task.info.state == 'success':
                logging.info(f"监听到exsi的任务执行成功")
                return "ok"

            if task.info.state == 'error':
                logging.error(f"监听到exsi的任务执行失败了")
                return task.info.error.msg

    # 获取指定虚拟机的信息
    def get_vm_info_once(self, vm_obj) -> dict:
        """
        获取所有虚拟机信息
        :param vm_obj: CreateContainerView获取的虚拟机对象
        """
        # 定义取哪些关键信息
        vm_properties = {
            "name_VM": "name",
            "uuid": "config.uuid",
            "CPU": "config.hardware.numCPU",
            "memory_MB": "config.hardware.memoryMB",
            "name_OS": "config.guestFullName",
            "status_tools": "guest.toolsRunningStatus",
            "IP": "guest.ipAddress",
            "status_power": "summary.runtime.powerState",
        }
        res = {}
        for k, v in vm_properties.items():
            # print(dir(vm_obj))
            # res[k] = getattr(vm_obj, "config.uuid")
            res[k] = self.get_attr(vm_obj, v)
        return res

    # 返回虚拟机指定的指标信息, 可指定名字也可以全要
    def get_vm_info_all(self, vm_name: str = None, vm_name_list: list = None) -> dict:
        """
        返回虚拟机指定的指标信息, 可指定名字也可以全要
        :param vm_name: 要查的虚拟机名字
        :param vm_name_list: 要查的虚拟机名字的列表
        """
        res = {}
        for vm in self.container_view.view:
            if vm_name:
                if vm.name != vm_name:
                    continue
            if vm_name_list:
                if vm.name not in vm_name_list:
                    continue
            res[vm.name] = self.get_vm_info_once(vm)
            # print(dir(vm.config.hardware))
        return res

    # 创建虚拟机
    def clone_vm(self, template_name: str, vm_name: str, vm_ip: str, vm_cpu: int = None, vm_member_mb: int = None):
        """
        :param template_name: 被克隆虚拟机的名字
        :param vm_name: 要新建的虚拟机的名字
        :param vm_ip: 要新建的虚拟机的IP
        :param vm_cpu: 要新建的虚拟机的cpu数量
        :param vm_member_mb: 要新建的虚拟机的内存大小, 单位: MB
        :param
        :param
        :return:
        """

        # 先定义这个被克隆的虚拟机实例为空
        template_vm = None
        template_n = 0
        # 寻找这个虚拟机
        for vm in self.container_view.view:
            if vm.name == template_name:
                template_vm = vm
                template_n += 1
        # 没找到就退出
        if not template_vm:
            return f"没找到名为 '{template_name}' 的虚拟机"
        # 如果获取到不止一个, 说明有重名的虚拟机, 立即退出报错
        if template_n != 1:
            return f"发现不止一个名为 '{template_name}' 的已存在虚拟机"

        # 获取这个模板的相关信息
        # template_info_tmp = self.get_vm_info_once(template_vm)
        # logging.info(template_info_tmp)
        template_info = self.get_vm_info_once(template_vm)
        # 如果没有传入某些配置, 那么保持和模板一致
        if not vm_cpu:
            vm_cpu = template_info["CPU"]
            logging.info(f"未指定cpu核数, 沿用模板虚拟机的: {vm_cpu}")
        if not vm_member_mb:
            vm_member_mb = template_info["memory_MB"]
            logging.info(f"未指定内存大小(MB), 沿用模板虚拟机的: {vm_member_mb}")

        # 获取当前所有虚拟机信息
        vm_info_all = self.get_vm_info_all()
        # 校验传入的某些参数是否和现有虚拟机冲突
        for k, v in vm_info_all.items():
            # 校验IP地址是否重复
            if vm_ip == v["IP"]:
                return f"指定的新IP:{vm_ip}与虚拟机:{v['name_VM']}的IP:{v['IP']}冲突"

        # 新建
        # 获取模版
        template = template_vm
        datacenter_name = "ha-datacenter"
        datastore_name = None
        cluster_name = None

        # 开始克隆
        clone_spec = vim.vm.CloneSpec()
        clone_spec.location = vim.vm.RelocateSpec()
        clone_spec.powerOn = False
        clone_spec.template = False

        # Clone the VM
        clone_task = template_vm.CloneVM_Task(folder=template_vm.parent, name=vm_name, spec=clone_spec)
        res = self.wait_for_task(clone_task)
        return res

if __name__ == '__main__':
    # 定义日志格式
    logging.basicConfig(
        level=logging.INFO,
        format="%(asctime)s %(levelname)s:%(filename)s:%(funcName)s:%(thread)d:%(lineno)d %(message)s"
    )
    a = exsi()
    # b = a.get_datastore_by_name("a", "ha-datacenter")
    b = a.clone_vm(
        template_name="template_centos7",
        vm_name="test-new-1",
        vm_ip="10.33.22.10"
    )
    logging.info(b)

Reproduction steps

1. directly run the script

Expected behavior

I hope to clone another virtual machine from an existing one

Additional context

No response

[prziborowski]

The CloneVm method only works from vCenter. You can't execute it directly from ESXi.

[snow-with-tea]

The CloneVm method only works from vCenter. You can't execute it directly from ESXi.

Is there any other way to achieve this effect? Creating one by one is very painful

[prziborowski]

If your template VM is very simple, then it is possible you could re-implement what clone method is doing (copying files and creating a new VM). I haven't done something like this in many years, so I don't know how well it still works. There is a VirtualDiskManager.CopyVirtualDisks

So from your esxi class, you'd call:

from pyVim.task import WaitForTask
task = self.content.virtualDiskManager.CopyVirtualDisks(sourceName=sourceName, destName=destName)
WaitForTask(task)

And once the disks are copied, you can use this create_vm.py sample: https://github.com/vmware/pyvmomi-community-samples/pull/523/files for an idea on attaching existing disks for a new VM. The other tricky part would be creating a ConfigSpec that looks similar to the ConfigInfo of the source VM, but replacing identifiers so that you have a unique VM. I don't have any pointers to any code that does that.

Alternatively I believe you can export the VM (as an ovf/ova), and import it to get a new VM. This won't be very performant though as it would be transferring disks through your client. For that, you would combine https://github.com/vmware/pyvmomi-community-samples/blob/master/samples/export_vm.py and https://github.com/vmware/pyvmomi-community-samples/blob/master/samples/deploy_ovf.py Or could leverage the ovftool binary that I believe Workstation and Fusion have.

If you are just using the same template over and over for the clones, you could host the ovf/vmdk on a webserver that ESXi has access to (might have to configure firewall rules for that) and the deployOvf/deployOva can be used to pull those files.

The easiest way is to use vCenter though.