x/vmware-nsx
Code Issues Proposed changes

NSX|V3: Change default value of metadata_on_demand to False

Browse Source 
Change the default value of metadata_on_demand from True to False
because the following reasons:

1. Due to current dnsmasq bug, new VMs on a DHCP-enabled subnet may still
   get the metadata route to DHCP port via dnsmasq DHCP options. Once the
   bug is fixed, new VMs will get the metadata route to router port via
   dnsmasq DHCP options. If we have metadata_on_demand=False and
   force_metadata=False, we can always have internal metadata network
   ready, which can handle the metadata requests routed to the router
   port if the dnsmasq bug is fixed.

2. According to current DHCP agent implementation, if DHCP agent is
   restarted, it will try to restart all metadata proxies. But it will
   skip the metadata proxy for a network that has any subnet attached
   to a router. Instead, DHCP agent will start a metadata-proxy for the
   router. If old metadata proxy processes are still running, then it
   should be fine. But consider the case when a openstack network node
   is restarted, then all old processes are gone. Thus DHCP agent will
   not start those metadata proxies for networks with attached router.
   This means any VM that has routing table containing a metadata route
   to the DHCP port will fail to reach metadata service because the
   corresponding metadata proxy that handle 169.254.169.254:80 is not
   running.

3. When (2) happens, if we have force_metadata=True, dnsmasq will
   provide metadata route to DHCP port for any new VM on a DHCP-enabled
   and router-attached subnet. So those VMs will fail to reach metadata
   service.

4. When (2) happens, if we have force_metadata=False, dnsmasq will
   provide metadata route to router port for any new VM on a DHCP-enabled
   and router-attached subnet. If metadata_on_demand=False, the
   pre-created internal metadata network can forward the metadata
   requests from those VMs. But if metadata_on_demand=True, the internal
   metadata network is not created because the router is attached to a
   DHCP-enabled subnet. Thus the router can not route those metadata
   requests.

Also fix metadata tags used in NSX|V3 unit tests.

Change-Id: I6d39dffa365f172ad24530ee938b5af3483a7a18
This commit is contained in:
Shih-Hao Li 2016-03-11 16:01:39 -08:00
parent c760363d83
commit 4a419f3790
3 changed files with 4 additions and 3 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
etc/nsx.ini
View File

@ -402,4 +402,4 @@
# If True, an internal metadata network will be created for a router only when
# the router is attached to a DHCP-disabled subnet.
# metadata_on_demand = True
# metadata_on_demand = False

2
vmware_nsx/common/config.py
View File

@ -253,7 +253,7 @@ nsx_v3_opts = [
"does not support namespaces otherwise access_network "
"should be used.")),
cfg.BoolOpt('metadata_on_demand',
default=True,
default=False,
help=_("If true, an internal metadata network will be created "
"for a router only when the router is attached to a "
"DHCP-disabled subnet.")),

3
vmware_nsx/tests/unit/nsx_v3/test_plugin.py
View File

@ -330,7 +330,8 @@ class TestL3NatTestCase(L3NatTest,
ext_mgr=None,
service_plugins=None):
super(TestL3NatTestCase, self).setUp(plugin=plugin, ext_mgr=ext_mgr)
cfg.CONF.set_override('metadata_mode', None, 'NSX')
cfg.CONF.set_override('metadata_mode', None, 'nsx_v3')
cfg.CONF.set_override('metadata_on_demand', False, 'nsx_v3')
def _test_create_l3_ext_network(
self, physical_network=nsx_v3_mocks.DEFAULT_TIER0_ROUTER_UUID):