nova network deallocates fixed IP with non unique field in DB

[willfu]

Tried many diff configurations for openstack network service.
Found Openstack deallocated fixd IP with non unique field in DB table
fixed_ips, whose field name is address.

How to reproduce:
    1.create openstack network 10.0.100.0/24, range 256.
    2.delete this network.
    3.recreate it.
    4.nova delete instance id.
Then the bug fired with exception as below.

2012-09-17 21:59:50 TRACE nova.rpc.amqp Traceback (most recent call last):
2012-09-17 21:59:50 TRACE nova.rpc.amqp File
"/usr/lib/python2.6/site-packages/nova/rpc/amqp.py", line 253, in _process_data
2012-09-17 21:59:50 TRACE nova.rpc.amqp rval = node_func(context=ctxt,
**node_args)
2012-09-17 21:59:50 TRACE nova.rpc.amqp File
"/usr/lib/python2.6/site-packages/nova/network/manager.py", line 257, in
wrapped
2012-09-17 21:59:50 TRACE nova.rpc.amqp return func(self, context, *args,
**kwargs)
2012-09-17 21:59:50 TRACE nova.rpc.amqp File
"/usr/lib/python2.6/site-packages/nova/network/manager.py", line 379, in
deallocate_for_instance
2012-09-17 21:59:50 TRACE nova.rpc.amqp super(FloatingIP,
self).deallocate_for_instance(context, **kwargs)
2012-09-17 21:59:50 TRACE nova.rpc.amqp File
"/usr/lib/python2.6/site-packages/nova/network/manager.py", line 257, in
wrapped
2012-09-17 21:59:50 TRACE nova.rpc.amqp return func(self, context, *args,
**kwargs)
2012-09-17 21:59:50 TRACE nova.rpc.amqp File
"/usr/lib/python2.6/site-packages/nova/network/manager.py", line 934, in
deallocate_for_instance
2012-09-17 21:59:50 TRACE nova.rpc.amqp **kwargs)
2012-09-17 21:59:50 TRACE nova.rpc.amqp File
"/usr/lib/python2.6/site-packages/nova/network/manager.py", line 230, in
deallocate_fixed_ip
2012-09-17 21:59:50 TRACE nova.rpc.amqp network =
self._get_network_by_id(context, fixed_ip['network_id'])
2012-09-17 21:59:50 TRACE nova.rpc.amqp File
"/usr/lib/python2.6/site-packages/nova/network/manager.py", line 1550, in
_get_network_by_id
2012-09-17 21:59:50 TRACE nova.rpc.amqp return self.db.network_get(context,
network_id)
2012-09-17 21:59:50 TRACE nova.rpc.amqp File
"/usr/lib/python2.6/site-packages/nova/db/api.py", line 755, in network_get
2012-09-17 21:59:50 TRACE nova.rpc.amqp return IMPL.network_get(context,
network_id)
2012-09-17 21:59:50 TRACE nova.rpc.amqp File
"/usr/lib/python2.6/site-packages/nova/db/sqlalchemy/api.py", line 120, in
wrapper
2012-09-17 21:59:50 TRACE nova.rpc.amqp return f(*args, **kwargs)
2012-09-17 21:59:50 TRACE nova.rpc.amqp File
"/usr/lib/python2.6/site-packages/nova/db/sqlalchemy/api.py", line 1933, in
network_get
2012-09-17 21:59:50 TRACE nova.rpc.amqp if not result:
2012-09-17 21:59:50 TRACE nova.rpc.amqp NetworkNotFound: Network 1 could not be
found.

[willfu (willfly0620)]

After walking through code, I got the coding solution for this bug.
in /usr/lib/python2.6/site-packages/nova/network/manager.py
    def deallocate_fixed_ip(self, context, address, host, **kwargs):
        """Call the superclass deallocate_fixed_ip if i'm the correct host
        otherwise call to the correct host"""
        fixed_ip = self.db.fixed_ip_get_by_address(context, address)
        network = self._get_network_by_id(context, fixed_ip['network_id'])

The fixed_ip_get_by_address() method should have more sanity check to avoid
this bug.

[Vish Ishaya (vishvananda)]

pretty sure this is a duplicate and has been fixed already.