n-d-r: Peering failing on mixed IPv4 + IPv6 updates

[Dr. Jens Harbott]

Scenario: Having BGP peers defined for IPv6 to advertise tenant networks, like for a standard deployment with public IPv6 connectivity. When a router has both IPv4 and IPv6 subnets attached, updates like adding a new IPv6 subnet also create updates for the IPv4 prefixes, but these crash the peering:

Jun 13 08:06:03 vm2 neutron-bgp-dragent[3280003]: netaddr.core.AddrFormatError: base address '2001:db8::308' is not IPv4
Jun 13 08:06:03 vm2 neutron-bgp-dragent[3280003]: During handling of the above exception, another exception occurred:
Jun 13 08:06:03 vm2 neutron-bgp-dragent[3280003]: Traceback (most recent call last):
Jun 13 08:06:03 vm2 neutron-bgp-dragent[3280003]: File "/usr/local/lib/python3.10/dist-packages/os_ken/lib/hub.py", line 69, in _launch
Jun 13 08:06:03 vm2 neutron-bgp-dragent[3280003]: return func(*args, **kwargs)
Jun 13 08:06:03 vm2 neutron-bgp-dragent[3280003]: File "/usr/local/lib/python3.10/dist-packages/os_ken/services/protocols/bgp/base.py", line 253, in start
Jun 13 08:06:03 vm2 neutron-bgp-dragent[3280003]: self._run(*args, **kwargs)
Jun 13 08:06:03 vm2 neutron-bgp-dragent[3280003]: File "/usr/local/lib/python3.10/dist-packages/os_ken/services/protocols/bgp/peer.py", line 683, in _run
Jun 13 08:06:03 vm2 neutron-bgp-dragent[3280003]: self._process_outgoing_msg_list()
Jun 13 08:06:03 vm2 neutron-bgp-dragent[3280003]: File "/usr/local/lib/python3.10/dist-packages/os_ken/services/protocols/bgp/peer.py", line 769, in _process_outgoing_msg_list
Jun 13 08:06:03 vm2 neutron-bgp-dragent[3280003]: self._send_outgoing_route(outgoing_msg)
Jun 13 08:06:03 vm2 neutron-bgp-dragent[3280003]: File "/usr/local/lib/python3.10/dist-packages/os_ken/services/protocols/bgp/peer.py", line 729, in _send_outgoing_route
Jun 13 08:06:03 vm2 neutron-bgp-dragent[3280003]: self._protocol.send(update_msg)
Jun 13 08:06:03 vm2 neutron-bgp-dragent[3280003]: File "/usr/local/lib/python3.10/dist-packages/os_ken/services/protocols/bgp/speaker.py", line 395, in send
Jun 13 08:06:03 vm2 neutron-bgp-dragent[3280003]: self._send_with_lock(msg)
Jun 13 08:06:03 vm2 neutron-bgp-dragent[3280003]: File "/usr/local/lib/python3.10/dist-packages/os_ken/services/protocols/bgp/speaker.py", line 384, in _send_with_lock
Jun 13 08:06:03 vm2 neutron-bgp-dragent[3280003]: self._socket.sendall(msg.serialize())
Jun 13 08:06:03 vm2 neutron-bgp-dragent[3280003]: File "/usr/local/lib/python3.10/dist-packages/os_ken/lib/packet/bgp.py", line 5245, in serialize
Jun 13 08:06:03 vm2 neutron-bgp-dragent[3280003]: tail = self.serialize_tail()
Jun 13 08:06:03 vm2 neutron-bgp-dragent[3280003]: File "/usr/local/lib/python3.10/dist-packages/os_ken/lib/packet/bgp.py", line 5465, in serialize_tail
Jun 13 08:06:03 vm2 neutron-bgp-dragent[3280003]: binpathattrs += pa.serialize()
Jun 13 08:06:03 vm2 neutron-bgp-dragent[3280003]: File "/usr/local/lib/python3.10/dist-packages/os_ken/lib/packet/bgp.py", line 3661, in serialize
Jun 13 08:06:03 vm2 neutron-bgp-dragent[3280003]: value = self.serialize_value()
Jun 13 08:06:03 vm2 neutron-bgp-dragent[3280003]: File "/usr/local/lib/python3.10/dist-packages/os_ken/lib/packet/bgp.py", line 3871, in serialize_value
Jun 13 08:06:03 vm2 neutron-bgp-dragent[3280003]: addrconv.ipv4.text_to_bin(self.value))
Jun 13 08:06:03 vm2 neutron-bgp-dragent[3280003]: File "/usr/local/lib/python3.10/dist-packages/os_ken/lib/addrconv.py", line 36, in text_to_bin
Jun 13 08:06:03 vm2 neutron-bgp-dragent[3280003]: ip = self._fallback(text, **self._addr_kwargs)
Jun 13 08:06:03 vm2 neutron-bgp-dragent[3280003]: File "/usr/local/lib/python3.10/dist-packages/netaddr/ip/__init__.py", line 930, in __init__
Jun 13 08:06:03 vm2 neutron-bgp-dragent[3280003]: value, prefixlen = parse_ip_network(_ipv4, addr,
Jun 13 08:06:03 vm2 neutron-bgp-dragent[3280003]: File "/usr/local/lib/python3.10/dist-packages/netaddr/ip/__init__.py", line 803, in parse_ip_network
Jun 13 08:06:03 vm2 neutron-bgp-dragent[3280003]: expanded_addr = _ipv4.expand_partial_address(val1)
Jun 13 08:06:03 vm2 neutron-bgp-dragent[3280003]: File "/usr/local/lib/python3.10/dist-packages/netaddr/strategy/ipv4.py", line 259, in expand_partial_address
Jun 13 08:06:03 vm2 neutron-bgp-dragent[3280003]: raise error
Jun 13 08:06:03 vm2 neutron-bgp-dragent[3280003]: netaddr.core.AddrFormatError: invalid partial IPv4 address: '2001:db8::308'!
Jun 13 08:06:03 vm2 neutron-bgp-dragent[3280003]: : netaddr.core.AddrFormatError: invalid partial IPv4 address: '2001:db8::308'!

Full logs attached.

[OpenStack Infra (hudson-openstack)]

Fix proposed to branch: master
Review: https://review.opendev.org/c/openstack/neutron-dynamic-routing/+/885993

[OpenStack Infra (hudson-openstack)]

Fix proposed to branch: master
Review: https://review.opendev.org/c/openstack/os-ken/+/885998

[Roberto Bartzen Acosta (rbartzen)]

I also see this issue in my local deployment.

Steps to reproduce:

Step 1 - default n-d-r config
----------------------------

# Provider network
openstack network create provider --external --provider-physical-network \
  provider --provider-network-type flat

# Address scope
openstack address scope create --share --ip-version 6 test-address-scope-ipv6

# Provider IPv6 Subnet pool
openstack subnet pool create --address-scope test-address-scope-ipv6 --pool-prefix 2001:db8:1:1::/64 --default-prefix-length 64 test-pool-ipv6-provider

# Provider IPv6 subnet
openstack subnet create --ip-version 6 --subnet-pool test-pool-ipv6-provider --network provider public-subnet-v6 --ipv6-address-mode dhcpv6-stateful --ipv6-ra-mode dhcpv6-stateful --allocation-pool start=2001:db8:1:1::4-2001:db8:1:1:ffff:ffff:ffff:ffff --gateway 2001:db8:1:1::1

# Provider IPv4 subnet - no address scope required...
openstack subnet create --ip-version 4 --network provider --dhcp --host-route destination=172.16.0.0/24,gateway=172.16.0.1 --subnet-range 172.16.0.0/24 public-subnet-v4

# self-service Network

openstack network create self-service

# Provider IPv6 Self-service Network
openstack subnet pool create --address-scope test-address-scope-ipv6 --share --default --pool-prefix 2001:db8:1234:8000::/50 --default-prefix-length 64 --max-prefix-length 64 test-default-pool-ipv6

openstack subnet create --ip-version 6 --subnet-pool test-default-pool-ipv6 --network self-service --ipv6-address-mode dhcpv6-stateful --ipv6-ra-mode dhcpv6-stateful self-service-v6

openstack subnet create --ip-version 4 --network self-service --dhcp --host-route destination=192.168.0.0/24,gateway=192.168.0.1 --subnet-range 192.168.0.0/24 self-service-v4

# create router
openstack router create router1

# add self-service subnet as an interface on the router
openstack router add subnet router1 self-service-v4
openstack router add subnet router1 self-service-v6

# Add the provider network as a gateway on each router.
openstack router set --external-gateway provider router1

# create bgp speaker
openstack bgp speaker create --ip-version 6 --local-as 65000 bgpspeaker
openstack bgp speaker add network bgpspeaker provider

# create a vm on the self-service network
openstack server create --image cirros --flavor 1vcpu --network=self-service --security-group cf2e7d53-0db7-4873-82ab-cf67eceda937 vm1

*** works without issues

----------------------------

Step 2 - aditional config to reproduce the problem

# create some random address scope for IPv4
openstack address scope create --ip-version 4 test-reproducer

# create the subnet linked with the IPv4 address scope
openstack subnet pool create --address-scope test-reproducer --pool-prefix 10.173.0.0/16 --default-prefix-length 16 self-service-v4-reproducer

# create an additional router
openstack router create router2

# add self-service subnet as an interface on the router ***** the self-service-v4-reproducer triggers the problem
openstack router add subnet router2 self-service-v4-reproducer
openstack router add subnet router2 self-service-v6

# Add the provider network as a gateway on each router.
openstack router set --external-gateway provid...

[Roberto Bartzen Acosta (rbartzen)]

Hi Frickler,

I fixed the problem with the mixed routes on n-d-r, can I submit a patch on this LP or should I open a new LP BUG?

Regards,
Roberto

[Roberto Bartzen Acosta (rbartzen)]

Proposed Patch: https://review.opendev.org/c/openstack/neutron-dynamic-routing/+/888787