def guest_state_report_engine(self):
"""
Guest 状态上报引擎
"""
guest_state_mapping = dict()
while True:
if Utils.exit_flag:
msg = 'Thread guest_state_report_engine say bye-bye'
print msg
logger.info(msg=msg)
return
try:
# 3 秒钟更新一次
time.sleep(config['engine_cycle_interval'] * 3)
threads_status['guest_state_report_engine'] = {
'timestamp': ji.Common.ts()
}
self.refresh_dom_mapping()
for uuid, dom in self.dom_mapping_by_uuid.items():
state = Guest.get_state(dom=dom)
if uuid in guest_state_mapping and guest_state_mapping[
uuid] == state:
continue
guest_state_mapping[uuid] = state
Guest.guest_state_report(dom=dom)
except:
log_emit.warn(traceback.format_exc())
def host_performance_collection_engine(self):
while True:
if Utils.exit_flag:
msg = 'Thread host_performance_collection_engine say bye-bye'
print msg
logger.info(msg=msg)
return
try:
time.sleep(config['engine_cycle_interval'])
threads_status['host_performance_collection_engine'] = {
'timestamp': ji.Common.ts()
}
self.ts = ji.Common.ts()
if self.ts % self.interval != 0:
continue
self.update_interfaces()
self.update_disks()
self.host_cpu_memory_performance_report()
self.host_traffic_performance_report()
self.host_disk_usage_io_performance_report()
except:
log_emit.warn(traceback.format_exc())
def update_ssh_key(dom=None, msg=None):
assert isinstance(dom, libvirt.virDomain)
assert isinstance(msg, dict)
if not dom.isActive():
log = u'欲更新 SSH-KEY 的目标虚拟机未处于活动状态。'
log_emit.warn(log)
return
from utils import QGA
libvirt_qemu.qemuAgentCommand(
dom,
json.dumps({
'execute': 'guest-exec',
'arguments': {
'path': 'mkdir',
'capture-output': False,
'arg': ['-p', '/root/.ssh']
}
}), 3, libvirt_qemu.VIR_DOMAIN_QEMU_AGENT_COMMAND_NOWAIT)
redirection_symbol = '>'
ret_s = list()
for i, ssh_key in enumerate(msg['ssh_keys']):
if i > 0:
redirection_symbol = '>>'
exec_ret = libvirt_qemu.qemuAgentCommand(
dom,
json.dumps({
'execute': 'guest-exec',
'arguments': {
'path':
'/bin/sh',
'capture-output':
True,
'arg': [
'-c', ' '.join([
'echo', '"' + ssh_key + '"',
redirection_symbol,
'/root/.ssh/authorized_keys'
])
]
}
}), 3, libvirt_qemu.VIR_DOMAIN_QEMU_AGENT_COMMAND_NOWAIT)
exec_ret = json.loads(exec_ret)
status_ret = QGA.get_guest_exec_status(
dom=dom, pid=exec_ret['return']['pid'])
# exec_ret_str = base64.b64decode(json.loads(status_ret)['return']['out-data'])
# ret_s.append(json.loads(exec_ret_str))
ret_s.append(status_ret)
return ret_s
def refresh_guest_state(self):
try:
self.refresh_dom_mapping()
for dom in self.dom_mapping_by_uuid.values():
Guest.guest_state_report(dom=dom)
except:
log_emit.warn(traceback.format_exc())
def guest_creating_progress_report_engine():
"""
Guest 创建进度上报引擎
"""
list_creating_guest = list()
template_size = dict()
while True:
if Utils.exit_flag:
msg = 'Thread guest_creating_progress_report_engine say bye-bye'
print msg
logger.info(msg=msg)
return
try:
try:
payload = q_creating_guest.get(
timeout=config['engine_cycle_interval'])
list_creating_guest.append(payload)
q_creating_guest.task_done()
except Queue.Empty as e:
pass
# 当有 Guest 被创建时，略微等待一下，避免复制模板的动作还没开始，就开始计算进度。这样会产生找不到镜像路径的异常。
time.sleep(1)
threads_status['guest_creating_progress_report_engine'] = {
'timestamp': ji.Common.ts()
}
for i, guest in enumerate(list_creating_guest):
template_path = guest['template_path']
storage = Storage(storage_mode=guest['storage_mode'],
dfs_volume=guest['dfs_volume'])
if template_path not in template_size:
template_size[template_path] = float(
storage.getsize(path=template_path))
system_image_size = storage.getsize(
path=guest['system_image_path'])
progress = system_image_size / template_size[template_path]
guest_event_emit.creating(uuid=guest['uuid'],
progress=int(progress * 90))
if progress >= 1:
del list_creating_guest[i]
except:
log_emit.warn(traceback.format_exc())
def clear_scene(self):
if self.dirty_scene:
self.dirty_scene = False
if self.guest.gf.exists(self.guest.system_image_path):
self.guest.gf.remove(self.guest.system_image_path)
else:
log = u'清理现场失败: 不存在的路径 --> ' + self.guest.guest_dir
logger.warn(msg=log)
log_emit.warn(msg=log)
def guest_state_report(dom=None):
try:
_uuid = dom.UUIDString()
state = Guest.get_state(dom=dom)
log = u' '.join([u'域', dom.name(), u', UUID', _uuid, u'的状态改变为'])
if state == GuestState.running.value:
log += u' Running。'
guest_event_emit.running(uuid=_uuid)
elif state == GuestState.booting.value:
log += u' Booting。'
guest_event_emit.booting(uuid=_uuid)
elif state == GuestState.blocked.value:
log += u' Blocked。'
guest_event_emit.blocked(uuid=_uuid)
elif state == GuestState.paused.value:
log += u' Paused。'
guest_event_emit.paused(uuid=_uuid)
elif state == GuestState.shutdown.value:
log += u' Shutdown。'
guest_event_emit.shutdown(uuid=_uuid)
elif state == GuestState.shutoff.value:
log += u' Shutoff。'
guest_event_emit.shutoff(uuid=_uuid)
elif state == GuestState.crashed.value:
log += u' Crashed。'
guest_event_emit.crashed(uuid=_uuid)
elif state == GuestState.pm_suspended.value:
log += u' PM_Suspended。'
guest_event_emit.pm_suspended(uuid=_uuid)
else:
log += u' NO_State。'
guest_event_emit.no_state(uuid=_uuid)
log_emit.info(msg=log)
except Exception as e:
log_emit.warn(e.message)
def host_state_report_engine(self):
"""
计算节点状态上报引擎
"""
# 首次启动时，做数据初始化
self.update_interfaces()
self.update_disks()
boot_time = ji.Common.ts()
while True:
if Utils.exit_flag:
msg = 'Thread host_state_report_engine say bye-bye'
print msg
logger.info(msg=msg)
return
try:
time.sleep(config['engine_cycle_interval'])
threads_status['host_state_report_engine'] = {
'timestamp': ji.Common.ts()
}
# 一分钟做一次更新
if ji.Common.ts() % 60 == 0:
self.update_interfaces()
self.update_disks()
host_event_emit.heartbeat(
message={
'node_id': self.node_id,
'cpu': self.cpu,
'cpuinfo': self.cpuinfo,
'memory': self.memory,
'dmidecode': self.dmidecode,
'interfaces': self.interfaces,
'disks': self.disks,
'system_load': os.getloadavg(),
'boot_time': boot_time,
'memory_available': psutil.virtual_memory().available,
'threads_status': threads_status,
'version': self.version
})
except:
log_emit.warn(traceback.format_exc())
def guest_performance_collection_engine(self):
while True:
if Utils.exit_flag:
msg = 'Thread guest_performance_collection_engine say bye-bye'
print msg
logger.info(msg=msg)
return
try:
time.sleep(config['engine_cycle_interval'])
threads_status['guest_performance_collection_engine'] = {
'timestamp': ji.Common.ts()
}
self.ts = ji.Common.ts()
if self.ts % self.interval != 0:
continue
if self.ts % 3600 == 0:
# 一小时做一次 垃圾回收 操作
for k, v in self.last_guest_cpu_time.items():
if (self.ts - v['timestamp']) > self.interval * 2:
del self.last_guest_cpu_time[k]
for k, v in self.last_guest_traffic.items():
if (self.ts - v['timestamp']) > self.interval * 2:
del self.last_guest_traffic[k]
for k, v in self.last_guest_disk_io.items():
if (self.ts - v['timestamp']) > self.interval * 2:
del self.last_guest_disk_io[k]
self.refresh_dom_mapping()
self.guest_cpu_memory_performance_report()
self.guest_traffic_performance_report()
self.guest_disk_io_performance_report()
except:
log_emit.warn(traceback.format_exc())
def instruction_process_engine(self):
self.init_conn()
ps = r.pubsub(ignore_subscribe_messages=False)
ps.subscribe(config['instruction_channel'])
while True:
if Utils.exit_flag:
msg = 'Thread instruction_process_engine say bye-bye'
print msg
logger.info(msg=msg)
return
threads_status['instruction_process_engine'] = dict()
threads_status['instruction_process_engine']['timestamp'] = ji.Common.ts()
# noinspection PyBroadException
try:
msg = ps.get_message(timeout=config['engine_cycle_interval'])
if msg is None or 'data' not in msg or not isinstance(msg['data'], basestring):
continue
try:
msg = json.loads(msg['data'])
if msg['action'] == 'pong':
continue
if msg['action'] == 'ping':
# 通过 ping pong 来刷存在感。因为经过实际测试发现，当订阅频道长时间没有数据来往，那么订阅者会被自动退出。
r.publish(config['instruction_channel'], message=json.dumps({'action': 'pong'}))
continue
except ValueError as e:
logger.error(e.message)
log_emit.error(e.message)
continue
if 'node_id' in msg and int(msg['node_id']) != self.node_id:
continue
# 下列语句繁琐写法如 <code>if 'action' not in msg or 'uuid' not in msg:</code>
if not all([key in msg for key in ['_object', 'action']]):
continue
extend_data = dict()
if msg['_object'] == 'guest':
self.refresh_guest_mapping()
if msg['action'] not in ['create']:
if msg['uuid'] not in self.guest_mapping_by_uuid:
if config['DEBUG']:
_log = u' '.join([u'uuid', msg['uuid'], u'在计算节点', self.hostname, u'中未找到.'])
logger.debug(_log)
log_emit.debug(_log)
raise RuntimeError('The uuid ' + msg['uuid'] + ' not found in current domains list.')
self.guest = self.guest_mapping_by_uuid[msg['uuid']]
if not isinstance(self.guest, libvirt.virDomain):
raise RuntimeError('Guest ' + msg['uuid'] + ' is not a domain.')
if msg['action'] == 'create':
t = threading.Thread(target=Guest.create, args=(self.conn, msg))
t.setDaemon(False)
t.start()
continue
elif msg['action'] == 'reboot':
if self.guest.reboot() != 0:
raise RuntimeError('Guest reboot failure.')
elif msg['action'] == 'force_reboot':
self.guest.destroy()
self.guest.create()
Guest.quota(guest=self.guest, msg=msg)
elif msg['action'] == 'shutdown':
if self.guest.shutdown() != 0:
raise RuntimeError('Guest shutdown failure.')
elif msg['action'] == 'force_shutdown':
if self.guest.destroy() != 0:
raise RuntimeError('Guest force shutdown failure.')
elif msg['action'] == 'boot':
if not self.guest.isActive():
if self.guest.create() != 0:
raise RuntimeError('Guest boot failure.')
Guest.quota(guest=self.guest, msg=msg)
elif msg['action'] == 'suspend':
if self.guest.suspend() != 0:
raise RuntimeError('Guest suspend failure.')
elif msg['action'] == 'resume':
if self.guest.resume() != 0:
raise RuntimeError('Guest resume failure.')
elif msg['action'] == 'delete':
root = ET.fromstring(self.guest.XMLDesc())
if self.guest.isActive():
self.guest.destroy()
self.guest.undefine()
system_disk = None
for _disk in root.findall('devices/disk'):
if 'vda' == _disk.find('target').get('dev'):
system_disk = _disk
if msg['storage_mode'] in [StorageMode.ceph.value, StorageMode.glusterfs.value]:
# 签出系统镜像路径
path_list = system_disk.find('source').attrib['name'].split('/')
if msg['storage_mode'] == StorageMode.glusterfs.value:
Guest.dfs_volume = path_list[0]
Guest.init_gfapi()
try:
Guest.gf.remove('/'.join(path_list[1:]))
except OSError:
pass
elif msg['storage_mode'] in [StorageMode.local.value, StorageMode.shared_mount.value]:
file_path = system_disk.find('source').attrib['file']
try:
os.remove(file_path)
except OSError:
pass
elif msg['action'] == 'reset_password':
if self.guest.setUserPassword(msg['user'], msg['password']) != 0:
raise RuntimeError('Guest reset password failure.')
elif msg['action'] == 'attach_disk':
if 'xml' not in msg:
_log = u'添加磁盘缺少 xml 参数'
raise KeyError(_log)
flags = libvirt.VIR_DOMAIN_AFFECT_CONFIG
if self.guest.isActive():
flags |= libvirt.VIR_DOMAIN_AFFECT_LIVE
# 添加磁盘成功返回时，ret值为0。可参考 Linux 命令返回值规范？
if self.guest.attachDeviceFlags(xml=msg['xml'], flags=flags) != 0:
raise RuntimeError('Attack disk failure.')
Guest.quota(guest=self.guest, msg=msg)
elif msg['action'] == 'detach_disk':
if 'xml' not in msg:
_log = u'分离磁盘缺少 xml 参数'
raise KeyError(_log)
flags = libvirt.VIR_DOMAIN_AFFECT_CONFIG
if self.guest.isActive():
flags |= libvirt.VIR_DOMAIN_AFFECT_LIVE
if self.guest.detachDeviceFlags(xml=msg['xml'], flags=flags) != 0:
raise RuntimeError('Detach disk failure.')
elif msg['action'] == 'update_ssh_key':
if not self.guest.isActive():
_log = u'欲更新 SSH-KEY 的目标虚拟机未处于活动状态。'
logger.warning(_log)
log_emit.warn(_log)
continue
ret = Guest.update_ssh_key(guest=self.guest, msg=msg)
logger.info(json.dumps(ret, ensure_ascii=False))
elif msg['action'] == 'allocate_bandwidth':
t = threading.Thread(target=Guest.allocate_bandwidth, args=(self.guest, msg))
t.setDaemon(False)
t.start()
continue
elif msg['action'] == 'adjust_ability':
t = threading.Thread(target=Guest.adjust_ability, args=(self.conn, self.guest, msg))
t.setDaemon(False)
t.start()
continue
elif msg['action'] == 'migrate':
# duri like qemu+ssh://destination_host/system
if 'duri' not in msg:
_log = u'迁移操作缺少 duri 参数'
raise KeyError(_log)
# https://rk4n.github.io/2016/08/10/qemu-post-copy-and-auto-converge-features/
flags = libvirt.VIR_MIGRATE_PERSIST_DEST | \
libvirt.VIR_MIGRATE_UNDEFINE_SOURCE | \
libvirt.VIR_MIGRATE_COMPRESSED | \
libvirt.VIR_MIGRATE_PEER2PEER | \
libvirt.VIR_MIGRATE_AUTO_CONVERGE
root = ET.fromstring(self.guest.XMLDesc())
if msg['storage_mode'] == StorageMode.local.value:
# 需要把磁盘存放路径加入到两边宿主机的存储池中
# 不然将会报 no storage pool with matching target path '/opt/Images' 错误
flags |= libvirt.VIR_MIGRATE_NON_SHARED_DISK
flags |= libvirt.VIR_MIGRATE_LIVE
if not self.guest.isActive():
_log = u'非共享存储不支持离线迁移。'
logger.error(_log)
log_emit.error(_log)
raise RuntimeError('Nonsupport online migrate with storage of non sharing mode.')
if self.init_ssh_client(hostname=msg['duri'].split('/')[2], user='******'):
for _disk in root.findall('devices/disk'):
_file_path = _disk.find('source').get('file')
disk_info = Disk.disk_info_by_local(image_path=_file_path)
disk_size = disk_info['virtual-size']
stdin, stdout, stderr = self.ssh_client.exec_command(
' '.join(['qemu-img', 'create', '-f', 'qcow2', _file_path, str(disk_size)]))
for line in stdout:
logger.info(line)
log_emit.info(line)
for line in stderr:
logger.error(line)
log_emit.error(line)
elif msg['storage_mode'] in [StorageMode.shared_mount.value, StorageMode.ceph.value,
StorageMode.glusterfs.value]:
if self.guest.isActive():
flags |= libvirt.VIR_MIGRATE_LIVE
flags |= libvirt.VIR_MIGRATE_TUNNELLED
else:
flags |= libvirt.VIR_MIGRATE_OFFLINE
if self.guest.migrateToURI(duri=msg['duri'], flags=flags) == 0:
if msg['storage_mode'] == StorageMode.local.value:
for _disk in root.findall('devices/disk'):
_file_path = _disk.find('source').get('file')
if _file_path is not None:
os.remove(_file_path)
else:
raise RuntimeError('Unknown storage mode.')
elif msg['_object'] == 'disk':
if msg['action'] == 'create':
if msg['storage_mode'] == StorageMode.glusterfs.value:
Guest.dfs_volume = msg['dfs_volume']
Guest.init_gfapi()
if not Disk.make_qemu_image_by_glusterfs(gf=Guest.gf, dfs_volume=msg['dfs_volume'],
image_path=msg['image_path'], size=msg['size']):
raise RuntimeError('Create disk failure with glusterfs.')
elif msg['storage_mode'] in [StorageMode.local.value, StorageMode.shared_mount.value]:
if not Disk.make_qemu_image_by_local(image_path=msg['image_path'], size=msg['size']):
raise RuntimeError('Create disk failure with local storage mode.')
elif msg['action'] == 'delete':
if msg['storage_mode'] == StorageMode.glusterfs.value:
Guest.dfs_volume = msg['dfs_volume']
Guest.init_gfapi()
if Disk.delete_qemu_image_by_glusterfs(gf=Guest.gf, image_path=msg['image_path']) \
is not None:
raise RuntimeError('Delete disk failure with glusterfs.')
elif msg['storage_mode'] in [StorageMode.local.value, StorageMode.shared_mount.value]:
if Disk.delete_qemu_image_by_local(image_path=msg['image_path']) is not None:
raise RuntimeError('Delete disk failure with local storage mode.')
elif msg['action'] == 'resize':
if 'size' not in msg:
_log = u'添加磁盘缺少 disk 或 disk["size"] 参数'
raise KeyError(_log)
used = False
if msg['guest_uuid'].__len__() == 36:
used = True
if used:
self.refresh_guest_mapping()
if msg['guest_uuid'] not in self.guest_mapping_by_uuid:
if config['DEBUG']:
_log = u' '.join([u'uuid', msg['uuid'], u'在计算节点', self.hostname, u'中未找到.'])
logger.debug(_log)
log_emit.debug(_log)
raise RuntimeError('Resize disk failure, because the uuid ' + msg['guest_uuid'] +
' not found in current domains.')
self.guest = self.guest_mapping_by_uuid[msg['guest_uuid']]
if not isinstance(self.guest, libvirt.virDomain):
raise RuntimeError('Resize disk failure, because the guest is not a domain.')
# 在线磁盘扩容
if used and self.guest.isActive():
if 'device_node' not in msg:
_log = u'添加磁盘缺少 disk 或 disk["device_node|size"] 参数'
raise KeyError(_log)
# 磁盘大小默认单位为KB，乘以两个 1024，使其单位达到GB
msg['size'] = int(msg['size']) * 1024 * 1024
if self.guest.blockResize(disk=msg['device_node'], size=msg['size']) != 0:
raise RuntimeError('Online resize disk failure in blockResize method.')
Guest.quota(guest=self.guest, msg=msg)
# 离线磁盘扩容
else:
if not all([key in msg for key in ['storage_mode', 'dfs_volume', 'image_path']]):
_log = u'添加磁盘缺少 disk 或 disk["storage_mode|dfs_volume|image_path|size"] 参数'
raise KeyError(_log)
if msg['storage_mode'] == StorageMode.glusterfs.value:
if not Disk.resize_qemu_image_by_glusterfs(dfs_volume=msg['dfs_volume'],
image_path=msg['image_path'],
size=msg['size']):
raise RuntimeError('Offline resize disk failure with glusterfs.')
elif msg['storage_mode'] in [StorageMode.local.value, StorageMode.shared_mount.value]:
if not Disk.resize_qemu_image_by_local(image_path=msg['image_path'], size=msg['size']):
raise RuntimeError('Offline resize disk failure with local storage mode.')
elif msg['action'] == 'quota':
self.refresh_guest_mapping()
if msg['guest_uuid'] not in self.guest_mapping_by_uuid:
if config['DEBUG']:
_log = u' '.join([u'uuid', msg['guest_uuid'], u'在计算节点', self.hostname, u'中未找到.'])
logger.debug(_log)
log_emit.debug(_log)
raise RuntimeError('Disk quota failure, because the uuid ' + msg['guest_uuid'] +
' not found in current domains.')
self.guest = self.guest_mapping_by_uuid[msg['guest_uuid']]
if not isinstance(self.guest, libvirt.virDomain):
raise RuntimeError('Disk quota failure, because the guest is not a domain.')
if not self.guest.isActive():
_log = u'磁盘 ' + msg['uuid'] + u' 所属虚拟机未处于活动状态。'
logger.warning(_log)
log_emit.warn(_log)
continue
Guest.quota(guest=self.guest, msg=msg)
elif msg['_object'] == 'snapshot':
self.refresh_guest_mapping()
if msg['uuid'] not in self.guest_mapping_by_uuid:
if config['DEBUG']:
_log = u' '.join([u'uuid', msg['uuid'], u'在计算节点', self.hostname, u'中未找到.'])
logger.debug(_log)
log_emit.debug(_log)
raise RuntimeError('Snapshot ' + msg['action'] + ' failure, because the uuid ' +
msg['uuid'] + ' not found in current domains.')
self.guest = self.guest_mapping_by_uuid[msg['uuid']]
if not isinstance(self.guest, libvirt.virDomain):
raise RuntimeError('Snapshot ' + msg['action'] + ' failure, because the guest is not a domain.')
if msg['action'] == 'create':
t = threading.Thread(target=Guest.create_snapshot, args=(self.guest, msg))
t.setDaemon(False)
t.start()
continue
elif msg['action'] == 'delete':
t = threading.Thread(target=Guest.delete_snapshot, args=(self.guest, msg))
t.setDaemon(False)
t.start()
continue
elif msg['action'] == 'revert':
t = threading.Thread(target=Guest.revert_snapshot, args=(self.guest, msg))
t.setDaemon(False)
t.start()
continue
elif msg['action'] == 'convert':
t = threading.Thread(target=Guest.convert_snapshot, args=(msg,))
t.setDaemon(False)
t.start()
continue
elif msg['_object'] == 'os_template_image':
if msg['action'] == 'delete':
if msg['storage_mode'] == StorageMode.glusterfs.value:
Guest.dfs_volume = msg['dfs_volume']
Guest.init_gfapi()
try:
Guest.gf.remove(msg['template_path'])
except OSError:
pass
elif msg['storage_mode'] in [StorageMode.local.value, StorageMode.shared_mount.value]:
try:
os.remove(msg['template_path'])
except OSError:
pass
elif msg['_object'] == 'global':
if msg['action'] == 'refresh_guest_state':
host_use_for_refresh_guest_state = Host()
t = threading.Thread(target=host_use_for_refresh_guest_state.refresh_guest_state, args=())
t.setDaemon(False)
t.start()
continue
else:
_log = u'未支持的 _object：' + msg['_object']
logger.error(_log)
log_emit.error(_log)
response_emit.success(_object=msg['_object'], action=msg['action'], uuid=msg['uuid'],
data=extend_data, passback_parameters=msg.get('passback_parameters'))
except redis.exceptions.ConnectionError as e:
logger.error(traceback.format_exc())
# 防止循环线程，在redis连接断开时，混水写入日志
time.sleep(5)
except:
# 防止循环线程，在redis连接断开时，混水写入日志
time.sleep(5)
logger.error(traceback.format_exc())
log_emit.error(traceback.format_exc())
response_emit.failure(_object=msg['_object'], action=msg.get('action'), uuid=msg.get('uuid'),
passback_parameters=msg.get('passback_parameters'))
def migrate(dom=None, msg=None):
assert isinstance(dom, libvirt.virDomain)
assert isinstance(msg, dict)
# https://rk4n.github.io/2016/08/10/qemu-post-copy-and-auto-converge-features/
flags = libvirt.VIR_MIGRATE_PERSIST_DEST | \
libvirt.VIR_MIGRATE_UNDEFINE_SOURCE | \
libvirt.VIR_MIGRATE_COMPRESSED | \
libvirt.VIR_MIGRATE_PEER2PEER | \
libvirt.VIR_MIGRATE_AUTO_CONVERGE
root = ET.fromstring(dom.XMLDesc())
if msg['storage_mode'] == StorageMode.local.value:
# 需要把磁盘存放路径加入到两边宿主机的存储池中
# 不然将会报 no storage pool with matching target path '/opt/Images' 错误
flags |= libvirt.VIR_MIGRATE_NON_SHARED_DISK
flags |= libvirt.VIR_MIGRATE_LIVE
if not dom.isActive():
err = u'非共享存储不支持离线迁移。'
log_emit.warn(err)
raise RuntimeError(
'Nonsupport offline migrate with storage of non sharing mode.'
)
ssh_client = Utils.ssh_client(hostname=msg['duri'].split('/')[2],
user='******')
for _disk in root.findall('devices/disk'):
_file_path = _disk.find('source').get('file')
disk_info = Storage.image_info_by_local(path=_file_path)
disk_size = disk_info['virtual-size']
stdin, stdout, stderr = ssh_client.exec_command(' '.join([
'qemu-img', 'create', '-f', 'qcow2', _file_path,
str(disk_size)
]))
for line in stdout:
log_emit.info(line)
for line in stderr:
log_emit.error(line)
elif msg['storage_mode'] in [
StorageMode.shared_mount.value, StorageMode.ceph.value,
StorageMode.glusterfs.value
]:
if dom.isActive():
flags |= libvirt.VIR_MIGRATE_LIVE
flags |= libvirt.VIR_MIGRATE_TUNNELLED
else:
flags |= libvirt.VIR_MIGRATE_OFFLINE
# duri like qemu+ssh://destination_host/system
if dom.migrateToURI(duri=msg['duri'], flags=flags) == 0:
if msg['storage_mode'] == StorageMode.local.value:
for _disk in root.findall('devices/disk'):
_file_path = _disk.find('source').get('file')
if _file_path is not None:
os.remove(_file_path)
else:
raise RuntimeError('Unknown storage mode.')
def instruction_process_engine(self):
ps = r.pubsub(ignore_subscribe_messages=False)
ps.subscribe(config['instruction_channel'])
while True:
if Utils.exit_flag:
msg = 'Thread instruction_process_engine say bye-bye'
print msg
logger.info(msg=msg)
return
threads_status['instruction_process_engine'] = {
'timestamp': ji.Common.ts()
}
msg = dict()
extend_data = dict()
try:
msg = ps.get_message(timeout=config['engine_cycle_interval'])
if msg is None or 'data' not in msg or not isinstance(
msg['data'], basestring):
continue
try:
msg = json.loads(msg['data'])
if msg['action'] == 'pong':
continue
if msg['action'] == 'ping':
# 通过 ping pong 来刷存在感。因为经过实际测试发现，当订阅频道长时间没有数据来往，那么订阅者会被自动退出。
r.publish(config['instruction_channel'],
message=json.dumps({'action': 'pong'}))
continue
except ValueError as e:
log_emit.error(e.message)
continue
if 'node_id' in msg and int(msg['node_id']) != self.node_id:
continue
# 下列语句繁琐写法如 <code>if '_object' not in msg or 'action' not in msg:</code>
if not all([key in msg for key in ['_object', 'action']]):
continue
logger.info(msg=msg)
if msg['_object'] == 'guest':
self.refresh_dom_mapping()
if msg['action'] not in ['create']:
self.dom = self.dom_mapping_by_uuid[msg['uuid']]
assert isinstance(self.dom, libvirt.virDomain)
if msg['action'] == 'create':
t = threading.Thread(target=Guest.create,
args=(self.conn, msg))
t.setDaemon(False)
t.start()
continue
elif msg['action'] == 'reboot':
Guest.reboot(dom=self.dom)
elif msg['action'] == 'force_reboot':
Guest.force_reboot(dom=self.dom, msg=msg)
elif msg['action'] == 'shutdown':
Guest.shutdown(dom=self.dom)
elif msg['action'] == 'force_shutdown':
Guest.force_shutdown(dom=self.dom)
elif msg['action'] == 'boot':
Guest.boot(dom=self.dom, msg=msg)
elif msg['action'] == 'suspend':
Guest.suspend(dom=self.dom)
elif msg['action'] == 'resume':
Guest.resume(dom=self.dom)
elif msg['action'] == 'delete':
Guest.delete(dom=self.dom, msg=msg)
elif msg['action'] == 'reset_password':
Guest.reset_password(dom=self.dom, msg=msg)
elif msg['action'] == 'attach_disk':
Guest.attach_disk(dom=self.dom, msg=msg)
elif msg['action'] == 'detach_disk':
Guest.detach_disk(dom=self.dom, msg=msg)
elif msg['action'] == 'update_ssh_key':
Guest.update_ssh_key(dom=self.dom, msg=msg)
elif msg['action'] == 'allocate_bandwidth':
t = threading.Thread(target=Guest.allocate_bandwidth,
args=(self.dom, msg))
t.setDaemon(False)
t.start()
continue
elif msg['action'] == 'adjust_ability':
t = threading.Thread(target=Guest.adjust_ability,
args=(self.dom, msg))
t.setDaemon(False)
t.start()
continue
elif msg['action'] == 'migrate':
Guest().migrate(dom=self.dom, msg=msg)
elif msg['_object'] == 'disk':
if msg['action'] == 'create':
Storage(storage_mode=msg['storage_mode'],
dfs_volume=msg['dfs_volume']).make_image(
path=msg['image_path'], size=msg['size'])
elif msg['action'] == 'delete':
Storage(storage_mode=msg['storage_mode'],
dfs_volume=msg['dfs_volume']).delete_image(
path=msg['image_path'])
elif msg['action'] == 'resize':
mounted = True if msg['guest_uuid'].__len__(
) == 36 else False
if mounted:
self.refresh_dom_mapping()
self.dom = self.dom_mapping_by_uuid[
msg['guest_uuid']]
# 在线磁盘扩容
if mounted and self.dom.isActive():
# 磁盘大小默认单位为KB，乘以两个 1024，使其单位达到 GiB
msg['size'] = int(msg['size']) * 1024 * 1024
# https://libvirt.org/html/libvirt-libvirt-domain.html#virDomainBlockResize
self.dom.blockResize(disk=msg['device_node'],
size=msg['size'])
Guest.quota(dom=self.dom, msg=msg)
# 离线磁盘扩容
else:
Storage(storage_mode=msg['storage_mode'],
dfs_volume=msg['dfs_volume']).resize_image(
path=msg['image_path'],
size=msg['size'])
elif msg['action'] == 'quota':
self.refresh_dom_mapping()
self.dom = self.dom_mapping_by_uuid[msg['guest_uuid']]
Guest.quota(dom=self.dom, msg=msg)
elif msg['_object'] == 'snapshot':
self.refresh_dom_mapping()
self.dom = self.dom_mapping_by_uuid[msg['uuid']]
if msg['action'] == 'create':
t = threading.Thread(target=Guest.create_snapshot,
args=(self.dom, msg))
t.setDaemon(False)
t.start()
continue
elif msg['action'] == 'delete':
t = threading.Thread(target=Guest.delete_snapshot,
args=(self.dom, msg))
t.setDaemon(False)
t.start()
continue
elif msg['action'] == 'revert':
t = threading.Thread(target=Guest.revert_snapshot,
args=(self.dom, msg))
t.setDaemon(False)
t.start()
continue
elif msg['action'] == 'convert':
t = threading.Thread(target=Guest.convert_snapshot,
args=(msg, ))
t.setDaemon(False)
t.start()
continue
elif msg['_object'] == 'os_template_image':
if msg['action'] == 'delete':
Storage(storage_mode=msg['storage_mode'],
dfs_volume=msg['dfs_volume']).delete_image(
path=msg['template_path'])
elif msg['_object'] == 'global':
if msg['action'] == 'refresh_guest_state':
t = threading.Thread(target=Host().refresh_guest_state,
args=())
t.setDaemon(False)
t.start()
continue
if msg['action'] == 'upgrade':
try:
log = self.upgrade(msg['url'])
log_emit.info(msg=log)
except subprocess.CalledProcessError as e:
log_emit.warn(e.output)
self.rollback()
continue
log = self.restart()
log_emit.info(msg=log)
if msg['action'] == 'restart':
log = self.restart()
log_emit.info(msg=log)
else:
err = u'未支持的 _object：' + msg['_object']
log_emit.error(err)
response_emit.success(
_object=msg['_object'],
action=msg['action'],
uuid=msg['uuid'],
data=extend_data,
passback_parameters=msg.get('passback_parameters'))
except KeyError as e:
log_emit.warn(e.message)
if msg['_object'] == 'guest':
if msg['action'] == 'delete':
response_emit.success(
_object=msg['_object'],
action=msg['action'],
uuid=msg['uuid'],
data=extend_data,
passback_parameters=msg.get('passback_parameters'))
except:
# 防止循环线程，在redis连接断开时，混水写入日志
time.sleep(5)
log_emit.error(traceback.format_exc())
response_emit.failure(
_object=msg['_object'],
action=msg.get('action'),
uuid=msg.get('uuid'),
passback_parameters=msg.get('passback_parameters'))