-
Bug
-
Resolution: Unresolved
-
Critical
-
None
-
None
-
13
-
False
-
-
False
-
-
-
-
FDP 24.E
When reside-on-redirect-chassis is set to true in the Logical Router Port connecting a vlan LS to a LR the NATed traffic is not properly working as the reply has not done the snat.
The topology to reproduce it is the next:
LS1 (provider flat network) <–> LR <–> LS2 (provider vlan network)
Where LS1 is:
_uuid : a788e636-0125-4137-a4a5-94c87be13aac
acls : []
copp : []
dns_records : [3d774817-c13c-46dc-9907-319e45c7f292]
external_ids : {"neutron:mtu"="1500", "neutron:network_name"=nova, "neutron:revision_number"="6"}
forwarding_groups : []
load_balancer : []
load_balancer_group : []
name : neutron-3cc4fa35-f05a-452d-a281-c3560a9aed22
other_config : {mcast_flood_unregistered="false", mcast_snoop="true", vlan-passthru="false"}
ports : [0cd69d3f-290b-4c1e-84a6-cc850f1a4ab7, 35861567-7cf6-4f8c-86de-33b13cbbaa76, 4b3d6b3d-ccf2-4d05-aa41-c90152e86977, 6a50ea44-f55e-45c1-b323-0da99be507a9, 7e947787-1dc0-427d-82b2-a9eebf0563a7, 80cdc012-deab-4604-9f2d-0bf6012de8ce, c0e31473-b09d-4f92-a4be-d287f34d8288]
LR is:
_uuid : 256f5ac8-7249-4f07-9584-a4caca8bdf2e copp : [] enabled : true external_ids : {"neutron:availability_zone_hints"="", "neutron:gw_port_id"="701872a3-d04f-46ca-bcee-9585fcdd3528", "neutron:revision_number"="11", "neutron:router_name"=geneve-router} load_balancer : [] load_balancer_group : [] name : neutron-a073cae8-664a-448a-8f55-38f4d2e1948d nat : [0d0c054e-8be0-4771-b214-5524bdeaf277, 40e51b68-5c3b-4875-bf25-81d96166c622, 99526592-eeed-4144-adbe-cb2f641470d3] options : {always_learn_from_arp_request="false", dynamic_neigh_routers="true"} policies : [] ports : [27a962ff-a60c-4d75-b6bb-6f6bab2a8258, 5c0725d4-e71e-49f1-9e5c-e779bd96ffac, f36904dc-28e2-4d9d-8f4f-e8751970af5b] static_routes : [6e6a55b8-d145-4f32-a906-6c8febb92563, d2a1e654-6954-45ad-b203-2c8196099af6]
And LS2 is:
_uuid : 6e12b2b3-e50a-464d-898f-fe05b250fff5
acls : []
copp : []
dns_records : [ff9597a2-7a0e-41e5-af0a-ceb2d6845f51]
external_ids : {"neutron:mtu"="1500", "neutron:network_name"=vlan-network-1, "neutron:revision_number"="2"}
forwarding_groups : []
load_balancer : []
load_balancer_group : []
name : neutron-4d5438e6-9cee-4725-9d2f-27e517b058c4
other_config : {mcast_flood_unregistered="false", mcast_snoop="true", vlan-passthru="false"}
ports : [4851108e-456b-458d-8d20-92d2c748ab14, 6f393f72-25d0-4b41-ae93-c8a504d85d53, 7143da09-d561-4894-8ffa-3c9a80245f77, ce790ef1-e4de-427b-9398-12a87e7bc03a]
In this sample, both LSs has localnet ports:
_uuid : 7143da09-d561-4894-8ffa-3c9a80245f77
addresses : [unknown]
dhcpv4_options : []
dhcpv6_options : []
dynamic_addresses : []
enabled : []
external_ids : {}
ha_chassis_group : []
name : provnet-0aa0dd62-cee7-4aca-9faa-f4a33f348efc
options : {mcast_flood="false", mcast_flood_reports="true", network_name=datacentre}
parent_name : []
port_security : []
tag : 1911
tag_request : []
type : localnet
up : false
_uuid : c0e31473-b09d-4f92-a4be-d287f34d8288
addresses : [unknown]
dhcpv4_options : []
dhcpv6_options : []
dynamic_addresses : []
enabled : []
external_ids : {}
ha_chassis_group : []
name : provnet-c6c7798c-4243-4e46-b970-572b5094579c
options : {mcast_flood="false", mcast_flood_reports="true", network_name=datacentre}
parent_name : []
port_security : []
tag : []
tag_request : []
type : localnet
up : false
And the logical router port connecting the LS2 to LR has reside-on-redirect-chassis=true
_uuid : f36904dc-28e2-4d9d-8f4f-e8751970af5b
enabled : []
external_ids : {"neutron:network_name"=neutron-4d5438e6-9cee-4725-9d2f-27e517b058c4, "neutron:revision_number"="3", "neutron:router_name"="a073cae8-664a-448a-8f55-38f4d2e1948d", "neutron:subnet_ids"="40812d41-12d0-4258-9446-fd0d91c09647"}
gateway_chassis : []
ha_chassis_group : []
ipv6_prefix : []
ipv6_ra_configs : {}
mac : "fa:16:3e:25:04:31"
name : lrp-4db4cf4a-5709-4d89-8f16-40a45445436f
networks : ["192.168.2.1/24"]
options : {reside-on-redirect-chassis="true"}
peer : []
Finally, the next is the VM with the NAT entry associated:
(overcloud) [stack@undercloud-0 ~]$ openstack server list +--------------------------------------+-------------+--------+------------------------------------------------+--------------+----------+ | ID | Name | Status | Networks | Image | Flavor | +--------------------------------------+-------------+--------+------------------------------------------------+--------------+----------+ | c245ecc4-efd4-4c8b-8a2e-1ea868d53c1e | test-vlan-1 | ACTIVE | vlan-network-1=192.168.2.236, 10.0.0.249 | cirros | m1.micro | _uuid : 6f393f72-25d0-4b41-ae93-c8a504d85d53 addresses : ["fa:16:3e:95:ee:66 192.168.2.236"] dhcpv4_options : d1f3fcc7-4388-4c80-8745-7e575b427272 dhcpv6_options : [] dynamic_addresses : [] enabled : true external_ids : {"neutron:cidrs"="192.168.2.236/24", "neutron:device_id"="c245ecc4-efd4-4c8b-8a2e-1ea868d53c1e", "neutron:device_owner"="compute:nova", "neutron:network_name"=neutron-4d5438e6-9cee-4725-9d2f-27e517b058c4, "neutron:port_fip"="10.0.0.249", "neutron:port_name"="", "neutron:project_id"=f7e961614dff4bc18bf68eb23b382ebd, "neutron:revision_number"="4", "neutron:security_group_ids"="aafffa92-5141-4373-83f4-377f8fdde97c"} ha_chassis_group : [] name : "dd6230ca-79f1-4307-b88e-fb185fc4e4ec" options : {mcast_flood_reports="true", requested-chassis=compute-0.redhat.local} parent_name : [] port_security : ["fa:16:3e:95:ee:66 192.168.2.236"] tag : [] tag_request : [] type : "" up : true
And the associated NAT entry:
_uuid : 99526592-eeed-4144-adbe-cb2f641470d3
allowed_ext_ips : []
exempted_ext_ips : []
external_ids : {"neutron:fip_external_mac"="fa:16:3e:75:3f:bb", "neutron:fip_id"="8e5da506-9cb7-4a06-934b-91d4b9d90f0c", "neutron:fip_network_
id"="3cc4fa35-f05a-452d-a281-c3560a9aed22", "neutron:fip_port_id"="dd6230ca-79f1-4307-b88e-fb185fc4e4ec", "neutron:revision_number"="26", "neutron:router_name"=neutron-a073cae8-664a-448a-8f55-38f4d2e1948d}
external_ip : "10.0.0.249"
external_mac : "fa:16:3e:75:3f:bb"
external_port_range : ""
logical_ip : "192.168.2.236"
logical_port : "dd6230ca-79f1-4307-b88e-fb185fc4e4ec"
options : {}
type : dnat_and_snat
With that, tcpdumping on the node where the VM is we get:
(ens5/br-ex) 09:21:19.309363 IP 10.0.0.52 > 10.0.0.249: ICMP echo request, id 22946, seq 1, length 64 (vm tap device) 09:21:19.310123 IP 10.0.0.52 > 192.168.2.236: ICMP echo request, id 22946, seq 1, length 64 (vm tap device) 09:21:19.310662 IP 192.168.2.236 > 10.0.0.52: ICMP echo reply, id 22946, seq 1, length 64 (br-ex device) 09:21:19.311033 ethertype IPv4, IP 192.168.2.236 > 10.0.0.52: ICMP echo reply, id 22946, seq 1, length 64
As it can be seen, the traffic leaves with the internal VM ip instead of having the snat applied to use 10.0.0.249 instead of 192.168.2.236
If we disable reside-on-redirect-chassis, then we get it properly (but then the non-NAT traffic will be broken as it will be tunneled (geneve) to the gateway node:
[root@compute-0 heat-admin]# tcpdump -ni ens5 icmp
dropped privs to tcpdump
tcpdump: verbose output suppressed, use -v or -vv for full protocol decode
listening on ens5, link-type EN10MB (Ethernet), capture size 262144 bytes
09:28:03.652485 IP 10.0.0.52 > 10.0.0.249: ICMP echo request, id 36165, seq 1, length 64
09:28:03.653071 IP 10.0.0.249 > 10.0.0.52: ICMP echo reply, id 36165, seq 1, length 64