L3 EVPN is broken with latest FRR 10.0.1 upgrade.

[sudhanshukumar22]

Description

This issue is raised from sonic where we upgraded to FRR 10.0.1 and found sonic-net/sonic-buildimage#21177 (comment)

L3 EVPN is broken with FRR 10.0.1 upgrade. Adding the below configuration and performing config reload/reboot results in vni in the vrf missing in configuration files and thus resulting in type 5 routes not getting programmed in the vrf

Sample configuration. Add this and perform config reload. The vni field will disappear after config reload.

vrf Vrf1
vni 100
exit-vrf
!
router bgp 65001
no bgp suppress-duplicates
neighbor 12.0.0.2 remote-as 65001
neighbor 13.0.0.2 remote-as 65001
!
address-family ipv4 unicast
network 11.0.0.1/32
exit-address-family
!
address-family l2vpn evpn
neighbor 12.0.0.2 activate
neighbor 13.0.0.2 activate
advertise-all-vni
advertise-svi-ip
exit-address-family
exit
!
router bgp 65001 vrf Vrf1
no bgp suppress-duplicates
bgp bestpath as-path multipath-relax
!
address-family ipv4 unicast
redistribute connected
exit-address-family
!
address-family l2vpn evpn
advertise ipv4 unicast
exit-address-family
exit

Version

10.0.1

How to reproduce

"VRF": {
"Vrf1": {
"vni": "500200"
}
},
"VXLAN_EVPN_NVO": {
"my-nvo": {
"source_vtep": "vtep101032"
}
},
"VXLAN_TUNNEL": {
"vtep101032": {
"src_ip": "10.1.0.32"
}
},
"VXLAN_TUNNEL_MAP": {
"vtep101032|map_50020_Vlan20": {
"vlan": "Vlan20",
"vni": "50020"
},
"vtep101032|map_76543_Vlan69": {
"vlan": "Vlan69",
"vni": "76543"
},
"vtep101032|map_500100_Vlan100": {
"vlan": "Vlan100",
"vni": "500100"
},
"vtep101032|map_500101_Vlan101": {
"vlan": "Vlan101",
"vni": "500101"
},
"vtep101032|map_500200_Vlan200": {
"vlan": "Vlan200",
"vni": "500200"
}
},

"LOOPBACK_INTERFACE": {
    "Loopback0": {},
    "Loopback0|10.1.0.32/32": {}
},

SONiC doesn't push FRR configuration when configuring maps. EVPN is configured in split mode. Here is the FRR configuration. Please add these configurations to the device and reboot. vni will not appear in under vrf of frr configuraiton

!
vrf Vrf1
vni 500200
exit-vrf
!
router bgp 65000
bgp router-id 10.1.0.32
bgp graceful-restart restart-time 300
bgp graceful-restart
bgp graceful-restart preserve-fw-state
neighbor 30.0.0.2 remote-as 65000
neighbor 40.0.0.3 remote-as 65000
!
address-family ipv4 unicast
network 10.1.0.32/32
exit-address-family
!
address-family l2vpn evpn
neighbor 30.0.0.2 activate
neighbor 40.0.0.3 activate
advertise-all-vni
advertise-svi-ip
vni 500100
rd 10.1.0.32:100
exit-vni
vni 500101
rd 10.1.0.32:101
exit-vni
vni 50020
rd 10.1.0.32:20
exit-vni
exit-address-family
exit
!
router bgp 65000 vrf Vrf1
!
address-family ipv4 unicast
redistribute connected
exit-address-family
!
address-family ipv6 unicast
redistribute connected
exit-address-family
!
address-family l2vpn evpn
advertise ipv4 unicast
advertise ipv6 unicast
rd 10.1.0.32:200
exit-address-family
exit
!

@donaldsharp @dgsudharsan
Donald was able to repro this issue on frr master as well. The issue here is that when below vrf config with VNI config is applied and saved in non-integrated frr mode, the zebra.conf does not have vni config.

vrf Vrf1
vni 500200
exit-vrf
!
This issue is not seen in frr integrated mode.
This needs to be followed up in frr community.
@sudhanshukumar22 , please file frr issue and link here.

Expected behavior

after config reload/reboot results in vni in the vrf configuration should be present and type 5 routes should be programmed in the vrf.

Actual behavior

after config reload/reboot results in vni in the vrf missing in configuration files and thus resulting in type 5 routes not getting programmed in the vrf

Additional context

This issue has been discussed with donald sharp @donaldsharp and found that the issue is present in latest FRR as well.

Checklist

• I have searched the open issues for this bug.
• I have not included sensitive information in this report.

[donaldsharp]

This is a problem with not using integrated configuration and mgmtd. Integrated configuration is the future all usage should be migrated to this.

[sudhanshukumar22]

@donaldsharp : Should we tell that all testing in sonic using FRR10.0.1 should use integrated mode only.