x/vmware-nsx
Code Issues Proposed changes

Fix IP recycling on exhausted pool

Browse Source 
When recycling an IP, allocation pools are loaded with an inner
join with ip avaiblity ranges. However, if the pool is exhausted,
there will be no availability ranges and the query will return
no results because of the inner join. It will therefore be
impossible to reuse IP addresses once the pool has been exhausted.

This patch tries to reload allocation pools without joining on
IP avaiability ranges if the first query did not return any result.
Replacing an inner join with an outer join will not be possible
as postgresql backend does not support 'SELECT ... FOR UPDATE' with
outer joins (for more info see bug 1215350).

The patch also contains unit tests for verifying IP are recycled
correctly even when the pool is exhausted.

Bug 1240353

Change-Id: I87f274bd759cb47be77717f0760d3704f9f34ade
This commit is contained in:
Salvatore Orlando 2013-10-16 03:14:16 -07:00
parent 1cb773f175
commit 10f4a40137
2 changed files with 54 additions and 14 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

18
neutron/db/db_base_plugin_v2.py
View File

@ -16,6 +16,7 @@
# under the License.
import datetime
import itertools
import random
import netaddr
@ -329,13 +330,22 @@ class NeutronDbPluginV2(neutron_plugin_base_v2.NeutronPluginBaseV2,
models_v2.IPAllocationPool).filter_by(subnet_id=subnet_id).
options(orm.joinedload('available_ranges', innerjoin=True)).
with_lockmode('update'))
# If there are no available ranges the previous query will return no
# results as it uses an inner join to avoid errors with the postgresql
# backend (see lp bug 1215350). In this case IP allocation pools must
# be loaded with a different query, which does not require lock for
# update as the allocation pools for a subnet are immutable.
# The 2nd query will be executed only if the first yields no results
unlocked_allocation_pools = (context.session.query(
models_v2.IPAllocationPool).filter_by(subnet_id=subnet_id))
# Find the allocation pool for the IP to recycle
pool_id = None
for allocation_pool in allocation_pools:
for allocation_pool in itertools.chain(allocation_pools,
unlocked_allocation_pools):
allocation_pool_range = netaddr.IPRange(
allocation_pool['first_ip'],
allocation_pool['last_ip'])
allocation_pool['first_ip'], allocation_pool['last_ip'])
if netaddr.IPAddress(ip_address) in allocation_pool_range:
pool_id = allocation_pool['id']
break
@ -451,7 +461,7 @@ class NeutronDbPluginV2(neutron_plugin_base_v2.NeutronPluginBaseV2,
for subnet in subnets:
range = range_qry.filter_by(subnet_id=subnet['id']).first()
if not range:
LOG.debug(_("All IP's from subnet %(subnet_id)s (%(cidr)s) "
LOG.debug(_("All IPs from subnet %(subnet_id)s (%(cidr)s) "
"allocated"),
{'subnet_id': subnet['id'], 'cidr': subnet['cidr']})
continue

50
neutron/tests/unit/test_db_plugin.py
View File

@ -22,6 +22,7 @@ import os
import random
import mock
import netaddr
from oslo.config import cfg
from testtools import matchers
import webob.exc
@ -1671,27 +1672,56 @@ fixed_ips=ip_address%%3D%s&fixed_ips=ip_address%%3D%s&fixed_ips=subnet_id%%3D%s
120)
self.assertTrue(log.mock_calls)
def test_recycle_ip_address_without_allocation_pool(self):
def _test_recycle_ip_address(self, ip_to_recycle, allocation_pools=None):
plugin = NeutronManager.get_plugin()
allocation_pools = [{"start": '10.0.0.10',
"end": '10.0.0.50'}]
if not allocation_pools:
allocation_pools = [{"start": '10.0.0.10',
"end": '10.0.0.50'}]
with self.subnet(cidr='10.0.0.0/24',
allocation_pools=allocation_pools) as subnet:
network_id = subnet['subnet']['network_id']
subnet_id = subnet['subnet']['id']
fixed_ips = [{"subnet_id": subnet_id,
"ip_address": '10.0.0.100'}]
"ip_address": ip_to_recycle}]
with self.port(subnet=subnet, fixed_ips=fixed_ips) as port:
update_context = context.Context('', port['port']['tenant_id'])
ctx = context.Context('', port['port']['tenant_id'])
ip_address = port['port']['fixed_ips'][0]['ip_address']
plugin._recycle_ip(update_context,
network_id,
subnet_id,
ip_address)
plugin._recycle_ip(ctx, network_id, subnet_id, ip_address)
q = update_context.session.query(models_v2.IPAllocation)
q = ctx.session.query(models_v2.IPAllocation)
q = q.filter_by(subnet_id=subnet_id)
self.assertEqual(q.count(), 0)
# If the IP address is in the allocation pool also verify the
# address is returned to the availability range
for allocation_pool in allocation_pools:
allocation_pool_range = netaddr.IPRange(
allocation_pool['start'], allocation_pool['end'])
if netaddr.IPAddress(ip_to_recycle) in allocation_pool_range:
# Do not worry about no result found exception
pool = ctx.session.query(
models_v2.IPAllocationPool).filter_by(
subnet_id=subnet_id).one()
ip_av_range = ctx.session.query(
models_v2.IPAvailabilityRange).filter_by(
allocation_pool_id=pool['id']).first()
self.assertIsNotNone(ip_av_range)
self.assertIn(netaddr.IPAddress(ip_to_recycle),
netaddr.IPRange(ip_av_range['first_ip'],
ip_av_range['last_ip']))
def test_recycle_ip_address_outside_allocation_pool(self):
self._test_recycle_ip_address('10.0.0.100')
def test_recycle_ip_address_in_allocation_pool(self):
self._test_recycle_ip_address('10.0.0.20')
def test_recycle_ip_address_on_exhausted_allocation_pool(self):
# Perform the recycle ip address on a subnet with a single address
# in the pool to verify the corner case exposed by bug 1240353
self._test_recycle_ip_address(
'10.0.0.20',
allocation_pools=[{'start': '10.0.0.20',
'end': '10.0.0.20'}])
def test_max_fixed_ips_exceeded(self):
with self.subnet(gateway_ip='10.0.0.3',