问题
OpenStack中很容易导致数据库和真实状态不一致的情况。因为OpenStack中操作基本都是分步完成的,从api接受请求到调度再到具体的操作节点,每一步都有可能更新数据库状态,如果哪一个出错就会直接抛出异常导致整个操作链中断,然后数据库就处于上一个操作后的更新状态。比较典型的就是删除实例,如果在nova-compute出错那这个实例的状态就可能永远处于deleting状态了。
现在我遇到这样一个问题,我有一个Volume挂载在一个实例上,但是不知道什么原因,这个Volume与这个实例的联系断了,在nova-volume通过tgtadm查看发现已经没有客户端连接到该Volume了。但是数据库中该记录还在,这导致以下结果:
1) 删除实例时无法删除,提示“Stderr: 'iscsiadm: No records found'”
2) 无法从实例卸载Volume 。于是只能直接操作数据库了。
与Volume相关的表
数据库中与Volume直接相关的几个表如下所示
操作Volume时数据库的相关数据变化
新建Volume:
select * from volumes where id = 40\G
*************************** 1. row ***************************
created_at: 2012-10-29 07:00:23
updated_at: 2012-10-29 07:00:25
deleted_at: NULL
deleted: 0
id: 40
ec2_id: NULL
user_id: 397dd3be88b6492caa88521502b07617
project_id: c6159a4f3dd34a2b83527499a40dbd2b
host: store2.sigsit.org
size: 20
availability_zone: nova
instance_id: NULL
mountpoint: NULL
attach_time: NULL
status: available
attach_status: detached
scheduled_at: 2012-10-29 07:00:23
launched_at: 2012-10-29 07:00:25
terminated_at: NULL
display_name: test
display_description:
provider_location: 10.61.2.14:3260,5 iqn.2010-10.org.openstack:volume-00000028 1
provider_auth: NULL
snapshot_id: NULL
volume_type_id: NULL
select * from volume_metadata where volume_id = 40\G
select * from iscsi_targets where volume_id = 40\G
*************************** 1. row ***************************
created_at: 2012-09-24 09:00:36
updated_at: 2012-10-29 07:00:24
deleted_at: NULL
deleted: 0
id: 205
target_num: 5
host: store2.sigsit.org
volume_id: 40
select * from block_device_mapping where volume_id = 40\G
select * from sm_volume where id = 40\G
将该Volume挂载到一个实例后:
select * from volumes where id = 40\G
*************************** 1. row ***************************
created_at: 2012-10-29 07:00:23
updated_at: 2012-10-29 11:55:36
deleted_at: NULL
deleted: 0
id: 40
ec2_id: NULL
user_id: 397dd3be88b6492caa88521502b07617
project_id: c6159a4f3dd34a2b83527499a40dbd2b
host: store2.sigsit.org
size: 20
availability_zone: nova
instance_id: 70
mountpoint: /dev/vdc
attach_time: NULL
status: in-use
attach_status: attached
scheduled_at: 2012-10-29 07:00:23
launched_at: 2012-10-29 07:00:25
terminated_at: NULL
display_name: test
display_description:
provider_location: 10.61.2.14:3260,5 iqn.2010-10.org.openstack:volume-00000028 1
provider_auth: NULL
snapshot_id: NULL
volume_type_id: NULL
select * from volume_metadata where volume_id = 40\G
select * from iscsi_targets where volume_id = 40\G
*************************** 1. row ***************************
created_at: 2012-09-24 09:00:36
updated_at: 2012-10-29 07:00:24
deleted_at: NULL
deleted: 0
id: 205
target_num: 5
host: store2.sigsit.org
volume_id: 40
select * from block_device_mapping where volume_id = 40\G
*************************** 1. row ***************************
created_at: 2012-10-29 11:55:36
updated_at: NULL
deleted_at: NULL
deleted: 0
id: 49
instance_id: 70
device_name: /dev/vdc
delete_on_termination: 0
virtual_name: NULL
snapshot_id: NULL
volume_id: 40
volu