一、問題現象

Cinder是H版程式碼，對接的儲存是Ceph。
使用cinder命令可以成功建立資料卷並備份資料卷。
但是，使用cinder backup-restore命令恢復備份卷時，如果指定了恢復到源卷，會導致源卷的狀態變為error_restoring：

# cinder backup-restore --volume-id <volume-id> <backup-id>

二、問題分析

在日誌/var/log/cinder/backup.log中，有如下報錯：
2014-06-26 15:53:26.456 28145 DEBUG cinder.backup.drivers.ceph [req-b873021e-b757-4d7d-be4c-235db3949930 admin 23770234f509405dbe5de13faf5bb976] dest volume is original volume - forcing full copy _diff_restore_allowed /usr/lib/python2.6/site-packages/cinder/backup/drivers/ceph.py:914
2014-06-26 15:53:26.643 28145 ERROR cinder.openstack.common.rpc.amqp [req-b873021e-b757-4d7d-be4c-235db3949930 admin 23770234f509405dbe5de13faf5bb976] Exception during message handling
2014-06-26 15:53:26.643 28145 TRACE cinder.openstack.common.rpc.amqp Traceback (most recent call last):
2014-06-26 15:53:26.643 28145 TRACE cinder.openstack.common.rpc.amqp   File "/usr/lib/python2.6/site-packages/cinder/openstack/common/rpc/amqp.py", line 441, in _process_data
2014-06-26 15:53:26.643 28145 TRACE cinder.openstack.common.rpc.amqp     **args)
2014-06-26 15:53:26.643 28145 TRACE cinder.openstack.common.rpc.amqp   File "/usr/lib/python2.6/site-packages/cinder/openstack/common/rpc/dispatcher.py", line 148, in dispatch
2014-06-26 15:53:26.643 28145 TRACE cinder.openstack.common.rpc.amqp     return getattr(proxyobj, method)(ctxt, **kwargs)
2014-06-26 15:53:26.643 28145 TRACE cinder.openstack.common.rpc.amqp   File "/usr/lib/python2.6/site-packages/cinder/utils.py", line 808, in wrapper
2014-06-26 15:53:26.643 28145 TRACE cinder.openstack.common.rpc.amqp     return func(self, *args, **kwargs)
2014-06-26 15:53:26.643 28145 TRACE cinder.openstack.common.rpc.amqp   File "/usr/lib/python2.6/site-packages/cinder/backup/manager.py", line 346, in restore_backup
2014-06-26 15:53:26.643 28145 TRACE cinder.openstack.common.rpc.amqp     {'status': 'available'})
2014-06-26 15:53:26.643 28145 TRACE cinder.openstack.common.rpc.amqp   File "/usr/lib64/python2.6/contextlib.py", line 23, in __exit__
2014-06-26 15:53:26.643 28145 TRACE cinder.openstack.common.rpc.amqp     self.gen.next()
2014-06-26 15:53:26.643 28145 TRACE cinder.openstack.common.rpc.amqp   File "/usr/lib/python2.6/site-packages/cinder/backup/manager.py", line 340, in restore_backup
2014-06-26 15:53:26.643 28145 TRACE cinder.openstack.common.rpc.amqp     backup_service)
2014-06-26 15:53:26.643 28145 TRACE cinder.openstack.common.rpc.amqp   File "/usr/lib/python2.6/site-packages/cinder/volume/drivers/rbd.py", line 800, in restore_backup
2014-06-26 15:53:26.643 28145 TRACE cinder.openstack.common.rpc.amqp     backup_service.restore(backup, volume['id'], rbd_fd)
2014-06-26 15:53:26.643 28145 TRACE cinder.openstack.common.rpc.amqp   File "/usr/lib/python2.6/site-packages/cinder/backup/drivers/ceph.py", line 986, in restore
2014-06-26 15:53:26.643 28145 TRACE cinder.openstack.common.rpc.amqp     self._restore_volume(backup, target_volume, volume_file)
2014-06-26 15:53:26.643 28145 TRACE cinder.openstack.common.rpc.amqp   File "/usr/lib/python2.6/site-packages/cinder/backup/drivers/ceph.py", line 976, in _restore_volume
2014-06-26 15:53:26.643 28145 TRACE cinder.openstack.common.rpc.amqp     volume_name, length, src_snap=restore_point)
2014-06-26 15:53:26.643 28145 TRACE cinder.openstack.common.rpc.amqp   File "/usr/lib/python2.6/site-packages/cinder/backup/drivers/ceph.py", line 775, in _full_restore
2014-06-26 15:53:26.643 28145 TRACE cinder.openstack.common.rpc.amqp     snapshot=src_snap, read_only=True)
2014-06-26 15:53:26.643 28145 TRACE cinder.openstack.common.rpc.amqp   File "/usr/lib/python2.6/site-packages/rbd.py", line 353, in __init__
2014-06-26 15:53:26.643 28145 TRACE cinder.openstack.common.rpc.amqp     raise make_ex(ret, 'error opening image %s at snapshot %s' % (name, snapshot))
2014-06-26 15:53:26.643 28145 TRACE cinder.openstack.common.rpc.amqp ImageNotFound: error opening image volume-1b04df30-2bdf-47d2-ad02-11ca22839790.backup.0b01d3ec-27d4-4d85-be51-197d4021b137 at snapshot None


根據上面的最後一行日誌，在cinder.backup.drivers.ceph:CephBackupDriver._full_restore中：
src_rbd = self.rbd.Image(client.ioctx, backup_name, snapshot=src_snap, read_only=True)
這行程式碼的snapshot=src_snap為None值，所以導致報錯。
往上回溯，snapshot=src_snap（即src_snap=restore_point）的值是在cinder.backup.drivers.ceph:CephBackupDriver._restore_volume中被賦值的：
diff_allowed, restore_point = self._diff_restore_allowed(base_name, backup, volume, volume_file, client)
我們進入到cinder.backup.drivers.ceph:CephBackupDriver._diff_restore_allowed中，看到這樣幾行程式碼：
        not_allowed = (False, None)


        # If the volume we are restoring to is the volume the backup was made
        # from, force a full restore since a diff will not work in this case.
        if volume['id'] == backup['volume_id']:
            LOG.debug(_("dest volume is original volume - forcing full copy"))
            return not_allowed

很明顯，參考註釋，這就是restore_point為什麼是None的原因。

三、解決辦法

據檢索，該問題和該連結中所描述的情況表現一致：
https://bugs.launchpad.net/cinder/+bug/1276977
該問題已經在Cinder的I版中修復了，修改程式碼如下：
https://git.openstack.org/cgit/openstack/cinder/commit/?id=4d1d9349fed815b37b0675d6f3936243a3909125