OpenStack Juno删除僵尸实例 - 虚机实例相关nova表

MQ服务消息超时或者任何其他中断实例创建或者删除的情况，都会造成“僵死”实例的存在。也即是这个实例并不存在，或者即便存在也是有问题的，并且在Dashboard仪表盘上删除不了。提示“you are not allowed to terminate this instance"之类的错误。即便提示删除成功，实际还是在那里死趟着。命令行同样的问题，如下

# nova list /*列表显示出来*/

+--------------------------------------+--------------------+--------+------------+-------------+---------------------+
| ID                                   | Name               | Status | Task State | Power State | Networks            |
+--------------------------------------+--------------------+--------+------------+-------------+---------------------+
| 09dca062-c577-469f-bff5-cccbe65a5b7c | instance_name      | ERROR  | deleting   | NOSTATE     |                     | 

# nova reset-state 09dca062-c577-469f-bff5-cccbe65a5b7c /*重置下状态，正在删除任务消失*/

# nova delete 09dca062-c577-469f-bff5-cccbe65a5b7c /*尝试命令行删除*/

# nova list /*还是上面的显示正在删除任务的状态*/

# mysql -uroot -p /*在控制节点上数据库去看看这个实例在哪个结算节点上*/

MariaDB [(none)]> select node from nova.instances where uuid='09dca062-c577-469f-bff5-cccbe65a5b7c'; /*得到这个实例所在的计算节点*/

# ll /var/lib/nova/instances/ /*到该计算节点查看实例文件夹在不在*/

# rm -Rf 09dca062-c577-469f-bff5-cccbe65a5b7c /*在的话，给删除*/

现在再次进入数据库，进行多个表格删除对应实例ID的记录。

1. nova.instances表中，实例ID的字段名是uuid；

2. 其他外链表使用的是instance_uuid来引用。

MariaDB [nova]> delete from table_name where instance_uuid='09dca062-c577-469f-bff5-cccbe65a5b7c'; /*从下面的表格中一一删除关联的记录*/

table_name如下：

1. security_group_instance_association，实例安全组，创建默认安全组是default，但不会记录，如果你已经关联了安全组，那就删除下*/

2. instance_info_caches，实例缓存，删除

3. block_device_mapping，实例块存储，默认保存了一条volume_id为NULL的记录，删除之*/

4. instance_actions_events，实例操作结果，字段action_id引用下表instance_actions中的id，该表没有instance_uuid*/

5. instance_actions，实例操作，每个操作有一个id，作为上表instance_actions_events的action_id*/

这里我在想，真是蛋疼，两个表为啥不结合为一个表。所以只能首先在instance_actions中获取操作id然后在events中再去删

MariaDB [nova]> select id,action,instance_uuid from instance_actions where instance_uuid='09dca062-c577-469f-bff5-cccbe65a5b7c';

+-----+--------+--------------------------------------+
| id  | action | instance_uuid                        |
+-----+--------+--------------------------------------+
| 380 | create | 09dca062-c577-469f-bff5-cccbe65a5b7c |
| 382 | delete | 09dca062-c577-469f-bff5-cccbe65a5b7c |
| 383 | delete | 09dca062-c577-469f-bff5-cccbe65a5b7c |
| 384 | delete | 09dca062-c577-469f-bff5-cccbe65a5b7c |
| 385 | delete | 09dca062-c577-469f-bff5-cccbe65a5b7c |
| 386 | delete | 09dca062-c577-469f-bff5-cccbe65a5b7c |
+-----+--------+--------------------------------------+

MariaDB [nova]> delete from instance_actions_events where action_id='380'; /*我在仪表盘操作了基础*/
MariaDB [nova]> delete from instance_actions_events where action_id='382';
MariaDB [nova]> delete from instance_actions_events where action_id='383';
MariaDB [nova]> delete from instance_actions_events where action_id='384';
MariaDB [nova]> delete from instance_actions_events where action_id='385';
MariaDB [nova]> delete from instance_actions_events where action_id='386'; /*删除instance_action_events之后再去删除instance_action*/

6. instance_faults，实例错误

7. instance_extra，没有看表的内容，不知道放的啥东西，只有一条记录

8. instance_system_metadata，实例元数据，记录蛮多的

9. instances，实例表，记录所有实例，仪表盘的实例列表就是读取这里

上面的9个表格删除完毕止之后，不要急着去刷新仪表盘，不然会提示无法获取实例列表信息 - 蛮吓人的！实例占用的资源依然被占用。

# systemctl restart openstack-nova-api openstack-nova-conductor /*重启下服务*/

# nova list /*实例删掉了，资源收回*/

最后如果每次都这么手动的去搞，麻烦，那就写个脚本：

# vim deletevm.sh

mysql -uroot -p << EOF
use nova;
delete from security_group_instance_association where instance_uuid='$1';
delete from instance_info_caches where instance_uuid='$1';

delete from block_device_mapping where instance_uuid='$1';

delete from instance_actions where instance_uuid='$1'; /*这个地方可能需要手工，SQL能力不强哎*/

delete from instance_faults where instance_uuid='$1';

delete from instance_extra where instance_uuid='$1';

delete from instance_system_metadata where instance_uuid='$1';

delete from instances where instance_uuid='$1';
EOF

# chmod +x deletevm.sh

# ./deletevm.sh instance_id /*以后用这个脚本来搞就快多了*/

# nova list /*最后可以再看看*/