[NETVIRT-1205] Neutron DHCP Port FIB entries are appearing in dual stack network Created: 10/Apr/18  Updated: 30/May/18  Resolved: 24/Apr/18

Status: Verified
Project: netvirt
Component/s: General
Affects Version/s: None
Fix Version/s: Fluorine

Type: Bug Priority: Medium
Reporter: Karthikeyan Krishnan Assignee: Karthikeyan Krishnan
Resolution: Won't Do Votes: 0
Labels: VPNV6
Remaining Estimate: 0 minutes
Time Spent: 1 week
Original Estimate: 1 week


 Description   

Neutron DHCP Port FIB entries are appearing in dual stack network when openstack q-dhcp service is being used.

cookie=0x8000003, duration=1837.161s, table=21, n_packets=0, n_bytes=0, priority=138,icmp6,metadata=0x30d40/0xfffffe,ipv6_dst=2001:db8:1111::1,icmp_type=128,icmp_code=0 actions=move:NXM _OF_ETH_SRC[]>NXM_OF_ETH_DST[],set_field:fa:16:3e:99:82:9b>eth_src,move:NXM_NX_IPV6_SRC[]>NXM_NX_IPV6_DST[],set_field:2001:db8:1111::1>ipv6_src,set_field:129->icmpv6_type,load:0->NXM_OF_IN_PORT[],resubmit(,21)
20 cookie=0x8000003, duration=1836.923s,
table=21, n_packets=0, n_bytes=0, priority=138,ipv6,metadata=0x30d40/0xfffffe,ipv6_dst=2001:db8:1111:0:f816:3eff:feac:5184 actions=group:150000
21 cookie=0x8000003, duration=1809.343s,
table=21, n_packets=0, n_bytes=0, priority=138,ipv6,metadata=0x30d40/0xfffffe,ipv6_dst=2001:db8:1111:0:f816:3eff:feee:500f actions=group:150001
22 cookie=0x8000003, duration=1802.736s,
table=21, n_packe
ts=0, n_bytes=0, priority=42,ip,metadata=0x30d40/0xfffffe,nw_dst=10.0.0.2 actions=group:150000

Where 10.0.0.2 is the dhcp-port and setup uses neutron dhcp-agent.
[vagrant@centos7_aione_ocata_odl SampleScripts]$ nova list
---------------------------------------------------------------------------------------------------------------------+

ID Name Status Task State Power State Networks

---------------------------------------------------------------------------------------------------------------------+

761838aa-35b9-4ad4-8466-bbf0b50c5917 vm1 ACTIVE
Running n1=10.0.0.5, 2001:db8:1111:0:f816:3eff:feee:500f

---------------------------------------------------------------------------------------------------------------------+
[vagrant@centos7_aione_ocata_odl SampleScripts]$ neutron port-list

[vagrant@centos7_aione_ocata_odl SampleScripts]$ neutron net-list
---------------------------------------------------------------------------------------------------------------------------------+

id name tenant_id subnets

---------------------------------------------------------------------------------------------------------------------------------+

34efd1ba-c5c8-4dd9-8065-f110ad7cf2f3 n1 ca9e7ba267d94cb486a6bca867b98e80 814bf8d7-1057-471c-9e30-56b267b53bfe 2001:db8:1111::/64
      7c6758ee-a2b6-4b6c-804b-ffdd98bb91aa 10.0.0.0/24

---------------------------------------------------------------------------------------------------------------------------------+
[vagrant@centos7_aione_ocata_odl SampleScripts]$
[vagrant@centos7_aione_ocata_odl SampleScripts]$ neutron router-port-list r1
--------------------------------------------------------------------------------------------------------------------------------------------------------

id name tenant_id mac_address fixed_ips

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

385cb961-9a67-46e5-868e-120c37e3a3a5   ca9e7ba267d94cb486a6bca867b98e80 fa:16:3e:99:82:9b {"subnet_id": "814bf8d7-1057-471c-9e30-56b267b53bfe", | | | | | | "ip_address": "2001:db8:1111::1"}

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

 Comments   
Comment by Karthikeyan Krishnan [ 24/Apr/18 ]

Based on the code review comments, dropping this fix. Since it may introduce some regression issue for external FLAT/VLAN based provider networks. Please refer the below link and comments for more details about the discussion.

https://git.opendaylight.org/gerrit/#/c/70764/

This is a use-case that is working today and changing it could introduce regressions. Tempest/Neutron functional/scenario tests do validate some use-cases where they try to ping to dhcp-port. I don't know if its within the network or across networks.

Either way, there could be users who may want to validate east-west/north-south use-case with dhcp-ports (instead of spawning new VMs). IMHO this is a reasonable expectation and is working today.

Worst case, I'm fine if we want to discard this patch and leave the functionality as-is. Its better than introducing some regression.

Generated at Wed Feb 07 20:23:29 UTC 2024 using Jira 8.20.10#820010-sha1:ace47f9899e9ee25d7157d59aa17ab06aee30d3d.