We provisioned an EVPN service between two PEs.

However, remote PE is unable to recognize peer as hosting a single-active EVPN service. This causes service impact.

Topology:
      MX480-1---|
CE1---|         MX960---CE2
      MX480-2---|

root@PE> show evpn instance LAB_EVPN_VLAN_300_FM extensive   
Aug 13 20:05:28
Instance: LAB_EVPN_VLAN_300_FM
  Route Distinguisher: x.y.z.16:8300
<snip>
  Number of ethernet segments: 2
    ESI: 00:01:10:10:11:10:00:00:00:00
      Status: Unresolved
      Number of remote PEs connected: 2
        Remote-PE        MAC-label  Aliasing-label  Mode
        x.y.z.11      0          0               single-active
        x.y.z.10      0          3401            single-homed    <<<<<

The function evpn_default_instance_esi_ad_rt_update sends both T1 and T4 route after esi bringup timer expired.

In the fix of PR1848102 (internal), it is optimized to avoid un-necessary DF role change when new IFL comes UP for the same ESI.

As the function evpn_default_instance_esi_ad_rt_update also triggers AD per ESI to update for all EVI RTs, which was missing and causing this issue.

It is fixed by triggering evpn_eolist_trigger_ad_route_per_esi when we skip calling evpn_eolist_bringup_start_timer. Added to skip the bringup timer ONLY for the case of EVPN_EDTI_GRANULARITY_PER_ESI.

This is fixed via internal PR 1863228. The

Workaround is to reset the interface facing CE at multi homed PE.

But when there are multiple IFLs in same IFD, resetting a single IFL can cause multiple IFLs going down.

External-TriggerThis issue might be seen if the following conditions are met:

*On all Junos and Junos OS Evolved Platforms

*multiple IFLs on the LAG interface

*'df-election-granularity per-esi' feature should be configured