This KB is only for customers using 21.4R3-S4.The customer reported AE interface was dropping all incoming packet after disabled a child port of AE in the VC member. This is Virtual Chassis and VXLAN+OVSDB environment, LAG is formed among different member switches. When disable a child port of AE in a VC member,
In this example, ae0 and ae1 is configured as OVSDB and they belongs to the same vlan. ae0 has ports ge-0/0/21 and ge-1/0/21 and ae1 has ports xe-0/0/24 and xe-1/0/24. When disabled ge-0/0/21 in ae0, ge-1/0/21 should forward the incoming packet to ae1 but ge-1/0/21 started dropping all incoming packets.
root@switch> show virtual-chassis Preprovisioned Virtual Chassis Virtual Chassis ID: 074c.e883.861a Virtual Chassis Mode: Enabled Mstr Mixed Route Neighbor List Member ID Status Serial No Model prio Role Mode Mode ID Interface 0 (FPC 0) Prsnt XH3120212345 qfx5120-48y-8c 129 Master* N VC 1 vcp-255/0/54 1 vcp-255/0/55 1 (FPC 1) Prsnt XH3120312345 qfx5120-48y-8c 129 Backup N VC 0 vcp-255/0/54 0 vcp-255/0/55 {master:0} root@switch> set protocols ovsdb interfaces ae0 set protocols ovsdb interfaces ae1 set vlans Contrail-00bd5e59-8da4-46c3-80b0-9f8665d79c6f interface ae0.2049 set vlans Contrail-00bd5e59-8da4-46c3-80b0-9f8665d79c6f interface ae1.2049 set vlans Contrail-00bd5e59-8da4-46c3-80b0-9f8665d79c6f vxlan vni 1066 root@d23-31> show vlans Contrail-00bd5e59-8da4-46c3-80b0-9f8665d79c6f Routing instance VLAN name Tag Interfaces default-switch Contrail-00bd5e59-8da4-46c3-80b0-9f8665d79c6f NA ae0.2049* ae1.2049* {master:0} root@switch> root@switch> show lacp interfaces ae0 Aggregated interface: ae0 LACP state: Role Exp Def Dist Col Syn Aggr Timeout Activity ge-0/0/21 Actor No No Yes Yes Yes Yes Fast Active ge-0/0/21 Partner No No Yes Yes Yes Yes Slow Active ge-1/0/21 Actor No No Yes Yes Yes Yes Fast Active ge-1/0/21 Partner No No Yes Yes Yes Yes Slow Active LACP protocol: Receive State Transmit State Mux State ge-0/0/21 Current Slow periodic Collecting distributing ge-1/0/21 Current Slow periodic Collecting distributing {master:0} root@switch>
When disabled ge-0/0/21, ge-1/0/21 starts dropping all incoming packets.
root@switch# set interfaces ge-0/0/21 disable {master:0}[edit] root@switch# commit root@switch# run show lacp interfaces ae0 Aggregated interface: ae0 LACP state: Role Exp Def Dist Col Syn Aggr Timeout Activity ge-0/0/21 Actor No Yes No No No Yes Fast Active ge-0/0/21 Partner No Yes No No No Yes Fast Passive ge-1/0/21 Actor No No Yes Yes Yes Yes Fast Active ge-1/0/21 Partner No No Yes Yes Yes Yes Slow Active LACP protocol: Receive State Transmit State Mux State ge-0/0/21 Port disabled No periodic Detached ge-1/0/21 Current Slow periodic Collecting distributing {master:0}[edit] root@switch#
Login vty in question. In this example, ge-1/0/21 is dropping so that we need to login to fpc1 from shell.
root@d23-31:RE:0% vty fpc1 FPC1(d23-31 vty)# show dcb ifd all ---------------------------------------------------------- ifd name global-dev local-dev port-num port-name ---------------------------------------------------------- et-1/0/48 1 0 87 xe44 ge-1/0/21 1 0 35 ge2 pfe-1/0/0 0 0 0 cpu0 pfh-1/0/0 0 0 0 cpu0 vcp-255/0/54 1 0 127 hg1 vcp-255/0/55 1 0 123 hg0 xe-1/0/24 1 0 44 xe23 FPC1(d23-31 vty)# You will see "discard" is set to All int eh output by the following command from shell. root@d23-31:RE:0% cprod -A fpc1 -c 'set dcbcm bcmshell "ps"' | egrep 'ge2|speed|port' ena/ speed/ link auto STP lrn inter max cut loop port link Lns duplex scan neg? state pause discrd ops face frame thru? back encap ge2( 35) up 1 1G FD HW No Forward All FA SGMII 9220 No IEEE <<<<< discrd is set to All. root@d23-31:RE:0%
Some fixes are using the brcm_vxlan_port_discard_set() API and the fix of PR1738404 is needed with these fixes but it was not included in only 21.4R3-S4. In VXLAN environment that AE (LAG) is formed among different member switches in Virtual Chassis, when disabled a child port of AE in a VC member, the same port numberon other member in the VC may drop all incoming packet. This is because non-local members are also requested for update in child member port discard settings. For example, AE0 has ports ge-0/0/1 and ge-1/0/1. When disable ge-0/0/1 in AE0, ge-1/0/1 should not discard the incoming packet but ge-1/0/1 may drop all incoming packets. This PR1738404 addresses this issue by validating the port/unit.