Description

In EVPN All-Active setup with evpn igmp-snooping configured for vlans, processing Type-8 ( Igmp leave synch) route by the multihoming peer PE device will cause BGP session flaps and evpn core isolation.

 

Symptoms

The Type-8 route NLRI will be considered as malformed and the BGP session will be flapped and leads to core isolation.
 

Mar 13 14:21:41.089 jtac-lab rpd[2245]: %DAEMON-4: bgp_rcv_nlri:11651: NOTIFICATION sent to 10.99.188.254 (Internal AS 198024): code 3 (Update Message Error) subcode 10 (bad address/prefix field), Reason: peer 10.99.188.254 (Internal AS 198024) UPDATE prefix length 16 too long

Mar 13 14:21:41.091 jtac-lab rpd[2245]: %DAEMON-3: Received malformed update from 10.99.188.254 (Internal AS 198024)

Mar 13 14:21:41.091 jtac-lab rpd[2245]: %DAEMON-3: Family evpn, prefix 8:10.99.188.254:1::9900040000000012::500::225.0.0.1::10.99.188.254:0/328

Mar 13 14:21:41.091 jtac-lab rpd[2245]: %DAEMON-4-RPD_BGP_NEIGHBOR_STATE_CHANGED: BGP peer 10.99.188.254 (Internal AS 198024) changed state from Established to Idle (event RecvUpdate) (instance master)

Mar 13 14:51:56.075 jtac-lab rpd[2245]: %DAEMON-3-BGP_EVPN_PFXLEN_EXCEEDS_MAXLEN: EVPN prefix length 2056 is longer than the longest NLRI 1992.

Mar 13 14:51:56.077 jtac-lab rpd[2245]: %DAEMON-4: bgp_rcv_nlri:11651: NOTIFICATION sent to 10.99.188.254 (Internal AS 198024): code 3 (Update Message Error) subcode 10 (bad address/prefix field), Reason: peer 10.99.188.254 (Internal AS 198024) UPDATE prefix length 2056 too long

Mar 13 14:51:56.081 jtac-lab rpd[2245]: %DAEMON-3: Received malformed update from 10.99.188.254 (Internal AS 198024)

Mar 13 14:51:56.083 jtac-lab rpd[2245]: %DAEMON-3: Family evpn, prefix 8:10.99.188.254:1::9900040000000012::500::225.0.0.1::10.99.188.254:0/328
 

Mar 13 14:25:08.019   jtac-lab-1 rpd[10341]: %DAEMON-4-RPD_BGP_NEIGHBOR_STATE_CHANGED: BGP peer 10.99.188.253 (Internal AS 198024) changed state from Established to Idle (event RecvNotify) (instance master)
Mar 13 14:25:08.019 jtac-lab-1 rpd[10341]: %DAEMON-3-EVPN_CORE_ISOLATED: EVPN core is isolated

Mar 13 14:55:22.961   jtac-lab-1 rpd[10341]: %DAEMON-4-RPD_BGP_NEIGHBOR_STATE_CHANGED: BGP peer 10.99.188.253 (Internal AS 198024) changed state from Established to Idle (event RecvNotify) (instance master)
Mar 13 14:55:22.962   jtac-lab-1 rpd[10341]: %DAEMON-3-EVPN_CORE_ISOLATED: EVPN core is isolated

Solution

This above behaviour is because of the addition of MRT (Maximum Response Time) field in the Type-8 routes in releases 21.3 and above to make it draft compatible.

The addition of MRT field will increase the NLRI length by 1 byte.
Release not supporting MRT field will decode the NLRI parameters incorrectly and drop the NLRI and triggers error message that cause BGP session flap.

Enable the following config knob in the device which supports MRT field in the type-8 route to make it interoperable between the release supporting MRT filed and older release which will not support MRT field in type-8 routes.

Config knob to be enabled

#set protocols evpn leave-sync-route-oldstyle 
 

Modification History

2024-04-01 : Article Created

Related Information

Reference doc

https://www.juniper.net/documentation/us/en/software/junos/cli-reference/topics/ref/statement/evpn-edit-routing-instances-protocols.html