openstack ice 生成虚拟机磁盘文件详解
感谢朋友支持本博客，欢迎共同探讨交流，由于能力和时间有限，错误之处在所难免，欢迎指正！
如有转载，请保留源作者博客信息。
如需交流，欢迎大家博客留言。
首先上图生成一个虚拟机在后台中的文件如何保存：
虚拟机生成选择类型表单：
生成成功后的虚拟机截图：
linux后台：
（1）、其中_base目录保存着虚拟机镜像模板文件（即cirros-0.3.2-x86_64镜像文件）：
格式为raw，即不可变文件，相对于qcow2性能更高，更多详细请自行查找相关资料
该文件也即前端页面的镜像：
比较纳闷的是为啥两个数据不一致（disk size: 17M ！= 13167616K），猜测是获取OS-EXT-IMG-SIZE:size时，做了一些处理。具体没仔细研究，有兴趣请自行研究。
（2）、其中一连串数字的目录（4184b91d-ef74-4c74-b88f-2f28048ec89b）下面保存了特定虚拟机自己的文件信息：
disk详情：
其中格式为qcow2,virtual size为前台flavor分配的root disk大小。disk size为目前实际占用的磁盘空间大小（因为qcow2是实际使用多少分配多少）
backing file：指明此虚拟机的后端文件，因此实例变化时候，只会改变差异信息到disk文件，_base下面的backing file镜像文件不会改变。
disk.info保存该实例的基本信息：
libvirt.xml启动虚拟机的整个配置文件：
以上讲解了整个实例创建完毕后，虚拟机实例文件是如何保存的。（注，本例中用的本地存储，假若用ceph作为后端，操作会有些许不同，请自行实验）
接下来对上述虚拟机文件的生成源码进行解析：
直接从virt/libvird/driver.py开始分析：
附上源码，详解：
|
def _create_image(self, context, instance,
disk_mapping, suffix='',
disk_images=None, network_info=None,
block_device_info=None, files=None,
admin_pass=None, inject_files=True):
import pydevd
#远程debug断点
pydevd.settrace('192.168.10.10', port=51234, stdoutToServer=True, stderrToServer=True)
if not suffix:
suffix = ''
booted_from_volume = self._is_booted_from_volume(
#判断是否为卷启动，本示例此处值为false，见图1-1
instance, disk_mapping)
def image(fname, image_type=CONF.libvirt.images_type):
return self.image_backend.image(instance,
fname + suffix, image_type)
def raw(fname):
return image(fname, image_type='raw')
# ensure directories exist and are writable
fileutils.ensure_tree(libvirt_utils.get_instance_path(instance))
#确认目录存在，且具有可写权限，变量值看图1-2
LOG.info(_('Creating image'), instance=instance)
# NOTE(dprince): for rescue console.log may already exist... chown it.
self._chown_console_log_for_instance(instance)
#控制台日志输出
# NOTE(yaguang): For evacuate disk.config already exist in shared
# storage, chown it.
self._chown_disk_config_for_instance(instance)
#修改disk保存路径的权限，具体参考图1-3
# NOTE(vish): No need add the suffix to console.log
libvirt_utils.write_to_file(
#写空字符串''到控制台路径文件中
self._get_console_log_path(instance), '', 7)
if not disk_images:
#获取一个disk%Ew��信息字典，本示例此处kernel_id，ramdisk_id都为空‘’
disk_images = {'image_id': instance['image_ref'],
'kernel_id': instance['kernel_id'],
'ramdisk_id': instance['ramdisk_id']}
if disk_images['kernel_id']:
fname = imagecache.get_cache_fname(disk_images, 'kernel_id')
raw('kernel').cache(fetch_func=libvirt_utils.fetch_image,
context=context,
filename=fname,
image_id=disk_images['kernel_id'],
user_id=instance['user_id'],
project_id=instance['project_id'])
if disk_images['ramdisk_id']:
fname = imagecache.get_cache_fname(disk_images, 'ramdisk_id')
raw('ramdisk').cache(fetch_func=libvirt_utils.fetch_image,
context=context,
filename=fname,
image_id=disk_images['ramdisk_id'],
user_id=instance['user_id'],
project_id=instance['project_id'])
inst_type = flavors.extract_flavor(instance)
#获取本次生成实例的类型即规格，见图1-4
# NOTE(ndipanov): Even if disk_mapping was passed in, which
# currently happens only on rescue - we still don't want to
# create a base image.
#是否从卷启动，不是从卷启动，则不需要创建base image，即共用_base文件夹下面镜像，作为后端文件backfile
if not booted_from_volume:
root_fname = imagecache.get_cache_fname(disk_images, 'image_id')#获取image_id的一个hash值
size = instance['root_gb'] * units.Gi
#获取页面选择的Root disk值
if size == 0 or suffix == '.rescue':
size = None
image('disk').cache(fetch_func=libvirt_utils.fetch_image,
#跟进：1.1
context=context,
filename=root_fname,
size=size,
image_id=disk_images['image_id'],
user_id=instance['user_id'],
project_id=instance['project_id'])
# Lookup the filesystem type if required
os_type_with_default = disk.get_fs_type_for_os_type(
#从instance参数中获取os_type，没有的话，则为default
instance['os_type'])
ephemeral_gb = instance['ephemeral_gb']
#获取页面的flavor中的ephemeral_gb 值
if 'disk.local' in disk_mapping:
#判断是否存在local分区，本示例不存在
disk_image = image('disk.local')
fn = functools.partial(self._create_ephemeral,
fs_label='ephemeral0',
os_type=instance["os_type"],
is_block_dev=disk_image.is_block_dev)
fname = "ephemeral_%s_%s" % (ephemeral_gb, os_type_with_default)
size = ephemeral_gb * units.Gi
disk_image.cache(fetch_func=fn,
filename=fname,
size=size,
ephemeral_size=ephemeral_gb)
for idx, eph in enumerate(driver.block_device_info_get_ephemerals(
block_device_info)):
disk_image = image(blockinfo.get_eph_disk(idx))
fn = functools.partial(self._create_ephemeral,
fs_label='ephemeral%d' % idx,
os_type=instance["os_type"],
is_block_dev=disk_image.is_block_dev)
size = eph['size'] * units.Gi
fname = "ephemeral_%s_%s" % (eph['size'], os_type_with_default)
disk_image.cache(
fetch_func=fn,
filename=fname,
size=size,
ephemeral_size=eph['size'])
if 'disk.swap' in disk_mapping:
#判断是否存在swap分区，本示例不存在
mapping = disk_mapping['disk.swap']
swap_mb = 0
swap = driver.block_device_info_get_swap(block_device_info)
if driver.swap_is_usable(swap):
swap_mb = swap['swap_size']
elif (inst_type['swap'] > 0 and
not block_device.volume_in_mapping(
mapping['dev'], block_device_info)):
swap_mb = inst_type['swap']
if swap_mb > 0:
size = swap_mb * units.Mi
image('disk.swap').cache(fetch_func=self._create_swap,
filename="swap_%s" % swap_mb,
size=size,
swap_mb=swap_mb)
# Config drive
if configdrive.required_by(instance):
#判断是否存在configdrive，本示例不存在。具体见图1-1-2：
LOG.info(_('Using config drive'), instance=instance)
extra_md = {}
if admin_pass:
extra_md['admin_pass'] = admin_pass
inst_md = instance_metadata.InstanceMetadata(instance,
content=files, extra_md=extra_md, network_info=network_info)
with configdrive.ConfigDriveBuilder(instance_md=inst_md) as cdb:
configdrive_path = self._get_disk_config_path(instance)
LOG.info(_('Creating config drive at %(path)s'),
{'path': configdrive_path}, instance=instance)
try:
cdb.make_drive(configdrive_path)
except processutils.ProcessExecutionError as e:
with excutils.save_and_reraise_exception():
LOG.error(_('Creating config drive failed '
'with error: %s'),
e, instance=instance)
# File injection only if needed
#如果inject_files ==True且inject_partition ！= -2则执行文件注入
elif inject_files and CONF.libvirt.inject_partition != -2:
if booted_from_volume:
LOG.warn(_('File injection into a boot from volume '
'instance is not supported'), instance=instance)
self._inject_data(
instance, network_info, admin_pass, files, suffix)
if%r0CONF.libvirt.virt_type == 'uml':
#如果配置文件配置为uml，则将disk的权限修改为root
libvirt_utils.chown(i-age('disk').path, 'root')
|
图1-1-2：判断是% 5��存在config_drive：
|
def required_by(instance):
return instance.get('config_drive') or CONF.force_config_drive
|
图1-1：
图1-2
图1-3：
图1-4：
1.1：image('disk').cache(fetch_func=libvirt_utils.fetch_image,
|
def cache(self, fetch_func, filename, size=None, *args, **kwargs):
"""Creates image from template.
#根据template创建image
Ensures that template and image not already exists.
Ensures that base directory exists.
Synchronizes on template fetching.
:fetch_func: Function that creates the base image
Should accept `target` argument.
:filename: Name of the file in the image directory
:size: Size of created image in bytes (optional)
"""
@utils.synchronized(filename, external=True, lock_path=self.lock_path)
def fetch_func_sync(target, *args, **kwargs):
fetch_func(target=target, *args, **kwargs)
base_dir = os.path.join(CONF.instances_path,
#获取base镜像的目录路径，见图1.1-1
CONF.image_cache_subdirectory_name)
if not os.path.exists(base_dir):
fileutils.ensure_tree(base_dir)
base = os.path.join(base_dir, filename)
#获取镜像文件路径，见图1.1-2
if not self.check_image_exists() or not os.path.exists(base):
self.create_image(fetch_func_sync, base, size,
#镜像和路径存在，创建镜像，跟进到1.1.1（此处为生成虚拟机文件核心代码）
*args, **kwargs)
if (size and self.preallocate and self._can_fallocate() and
os.access(self.path, os.W_OK)):
utils.execute('fallocate', '-n', '-l', size, self.path)
#linux命令：为文件预分配物理空间，本示例此处代码不执行
|
注：上述的
create_image函数拼装执行一个qemu-img create命令，最后一行代码则拼装执行一个
fallocate命令。
图1.1-1
图1.1-2
1.1.1：def create_image(self, prepare_template, base, size, *args, **kwargs):
在imagebackend.py文件中有四个create_image函数存在：
虚拟机镜像后端配置backend：
通过类和名字可以知道，不同的配置调用不同的create_image函数。Lvm、Qcow2、Raw、Rbd都继承了Image类，实现了对应的create_image函数。本示例中执行的为Qcow2类中create_image函数。（注：ceph对应为Rbd）
|
def create_image(self, prepare_template, base, size, *args, **kwargs):
filename = os.path.split(base)[-1]
#获取镜像文件名
@utils.synchronized(filename, external=True, lock_path=self.lock_path)
def copy_qcow2_image(base, target, size):
# TODO(pbrady): Consider copying the cow image here
# with preallocation=metadata set for performance reasons.
# This would be keyed on a 'preallocate_images' setting.
libvirt_utils.create_cow_image(base, target)
if size:
disk.extend(target, size, use_cow=True)
# Download the unmodified base image unless we already have a copy.
if not os.path.exists(base):
prepare_template(target=base, max_size=size, *args, **kwargs)
else:
#验证base的size是否满足小于size（其中base的size为'/var/lib/nova/instances/_base/70ec07731557ccdb9287d902cf15ec3d26ab9b41'文件的大小，第二个size为flovar中的root disk大小即1GB），如果不满足则说明，该镜像不能在小于自己容量的flaver中生成，报错
self.verify_base_size(base, size)
legacy_backing_size = None
legacy_base = base
#以下针对qcow2格式disk进行逻辑处理，因为qcow2文件占用空间是按需动态增长的
# Determine whether an existing qcow2 disk uses a legacy backing by
# actually looking at the image itself and parsing the output of the
# backing file it expects to be using.
if os.path.exists(self.path):
#判断路径是否存在，见图1.1.1-1
backing_path = libvirt_utils.get_disk_backing_file(self.path)
if backing_path is not None:
backing_file = os.path.basename(backing_path)
backing_parts = backing_file.rpartition('_')
if backing_file != backing_parts[-1] and \
backing_parts[-1].isdigit():
legacy_backing_size = int(backing_parts[-1])
legacy_base += '_%d' % legacy_backing_size
legacy_backing_size *= units.Gi
# Create the legacy backing file if necessary.
if legacy_backing_size:
if not os.path.exists(legacy_base):
with fileutils.remove_path_on_error(legacy_base):
libvirt_utils.copy_image(base, legacy_base)#跟进代码1.1.2
disk.extend(legacy_base, legacy_backing_size, use_cow=True)
if not os.path.exists(self.path):
#代码同上，同样可以参考图1.1.1-1
with fileutils.remove_path_on_error(self.path):
copy_qcow2_image(base, self.path, size)
#此处跟进代码1.1.4
|
图1.1.1-1：
1.1.2:copy_image(此函数作用为使用cp或者scp或者rsync命令复制disk image文件到指定目录)
1.1.4：def copy_qcow2_image(base, target, size):
|
@utils.synchronized(filename, external=True, lock_path=self.lock_path)
def copy_qcow2_image(base, target, size):
# TODO(pbrady): Consider copying the cow image here
# with preallocation=metadata set for performance reasons.
# This would be keyed on a 'preallocate_images' setting.
libvirt_utils.create_cow_image(base, target)
#跟进到1.1.5：
if size:
disk.extend(target, size, use_cow=True)
|
1.1.5：libvirt_utils.create_cow_image(base, target)
|
def create_cow_image(backing_file, path, size=None):
#拼装qemu-img 创建命令
"""Create COW image Creates a COW image with the given backing file :param backing_file: Existing image on which to base the COW image :param path: Desired location of the COW image """ base_cmd = ['qemu-img', 'create', '-f', 'qcow2'] cow_opts = [] if backing_file: cow_opts += ['backing_file=%s' % backing_file] base_details = images.qemu_img_info(backing_file) else: base_details = None # This doesn't seem to get inherited so force it to... # http://paste.ubuntu.com/1213295/ # TODO(harlowja) probably file a bug against qemu-img/qemu if base_details and base_details.cluster_size is not None: cow_opts += ['cluster_size=%s' % base_details.cluster_size] # For now don't inherit this due the following discussion... # See: http://www.gossamer-threads.com/lists/openstack/dev/10592 # if 'preallocation' in base_details: # cow_opts += ['preallocation=%s' % base_details['preallocation']] if base_details and base_details.encryption: cow_opts += ['encryption=%s' % base_details.encryption] if size is not None: cow_opts += ['size=%s' % size] if cow_opts: # Format as a comma separated list csv_opts = ",".join(cow_opts) cow_opts = ['-o', csv_opts] cmd = base_cmd + cow_opts + [path] execute(*cmd) #直接拼装出来的命令，本示例具体值，见图1.1.5-1（此处即拼装处生成虚拟机文件的命令） |
注：
|
如果“-o”选项中使用了backing_file这个选项来指定其后端镜像文件，那么这个创建的镜像文件仅记录与后端镜像文件的差异部分。后端镜像文件不会被修改，除非在QEMU monitor中使用“commit”命令或者使用“qemu-img commit”命令去手动提交这些改动。这种情况下，size参数不是必须需的，其值默认为后端镜像文件的大小。另外，直接使用“-b backfile”参数也与“-o backing_file=backfile”效果相同。 size选项用于指定镜像文件的大小，其默认单位是字节（bytes），也可以支持k（或K）、M、G、T来分别表示KB、MB、GB、TB大小。另外，镜像文件的大小（size）也并非必须写在命令的最后，它也可以被写在“-o”选项中作为其中一个选项。 |
上述命令执行完在生成disk文件：
跟进讲解到此结束。后续生成xml文件，请自行参考其他博文，此处不再重复讲解。