-
Bug
-
Resolution: Done-Errata
-
Major
-
4.12.0
-
No
-
1
-
SDN Sprint 253
-
1
-
False
-
Description of problem:
When adding another IP address to br-ex, geneve traffic sent from this node may be sent with the new IP address rather than the one configured for this tunnel. This will cause traffic to be dropped by the destination with the error: [root@ovn-control-plane openvswitch]# cat ovs-vswitchd.log | grep fc00:f853:ccd:e793::4 2024-04-17T16:47:02.146Z|00012|tunnel(revalidator10)|WARN|receive tunnel port not found (tcp6,tun_id=0xff0003,tun_src=0.0.0.0,tun_dst=0.0.0.0,tun_ipv6_src=fc00:f853:ccd:e793:ffff::1,tun_ipv6_dst=fc00:f853:ccd:e793::3,tun_gbp_id=0,tun_gbp_flags=0,tun_tos=0,tun_ttl=64,tun_erspan_ver=0,gtpu_flags=0,gtpu_msgtype=0,tun_flags=csum|key,in_port=5,vlan_tci=0x0000,dl_src=0a:58:2b:22:eb:86,dl_dst=0a:58:92:3f:71:e5,ipv6_src=fc00:f853:ccd:e793::4,ipv6_dst=fd00:10:244:1::7,ipv6_label=0x630b1,nw_tos=0,nw_ecn=0,nw_ttl=63,nw_frag=no,tp_src=8080,tp_dst=59130,tcp_flags=syn|ack) This is more likely to occur on ipv6 than ipv4, due to IP address ordering on the NIC and linux rules used to determine source IP to use when sending host originated traffic.
Version-Release number of selected component (if applicable):
All versions
How reproducible:
Always
To workaround with ipv6, set preferred_lft 0 on the address, which will cause it to become deprecated and linux will choose an alternative. Alternatively set external_ids:ovn-set-local-ip="true" in openvswitch on each node, which will force OVN to use the configured geneve-encap-ip. Related OVN issue: https://issues.redhat.com/browse/FDP-570
- is caused by
-
FDP-570 ovn-encap-ip not respected for geneve tunnels
- Closed
- links to
-
RHEA-2024:0041 OpenShift Container Platform 4.16.z bug fix update