Contrail BGP : Route update ceases until reboot of the node

[Deepak Jeyaraman]

Contrail 2.20 + Icehouse

Initially, when the control and compute nodes were provisioned, route updates via BGP was fine. After 3-4 days and instantiating 300 instances (over the period) and deletes of instances , BGP stuck in a weird state where the updates were only being sent for the 1st route target .

After reboot of the control node, things were fine:

MX output:

left-network-0.inet.0: 10 destinations, 10 routes (10 active, 0 holddown, 0 hidden)
  Prefix Nexthop MED Lclpref AS path
* 10.0.0.3/32 77.0.0.2 100 ?
* 10.0.1.3/32 77.0.0.2 100 ?
* 13.0.2.0/24 77.0.0.2 100 ?
* 14.0.0.0/24 77.0.0.2 100 ?
* 15.0.0.0/24 77.0.0.2 100 ?
* 192.168.4.3/32 77.0.0.2 100 ?
* 192.168.4.4/32 77.0.0.2 100 ?

left-network-1.inet.0: 10 destinations, 10 routes (10 active, 0 holddown, 0 hidden)
  Prefix Nexthop MED Lclpref AS path
* 10.0.0.3/32 77.0.0.2 100 ?
* 10.0.1.3/32 77.0.0.2 100 ?
* 13.0.1.0/24 77.0.0.2 100 ?
* 14.0.0.0/24 77.0.0.2 100 ?
* 15.0.0.0/24 77.0.0.2 100 ?
* 192.168.4.3/32 77.0.0.2 100 ?
* 192.168.4.4/32 77.0.0.2 100 ?

Issue:
MX not receiving route updates

left-network-0.inet.0: 7 destinations, 7 routes (7 active, 0 holddown, 0 hidden)
  Prefix Nexthop MED Lclpref AS path
* 10.0.0.3/32 77.0.0.2 200 ?
* 14.0.0.0/24 77.0.0.2 200 ?
* 15.0.0.0/24 77.0.0.2 200 ?
* 192.168.4.3/32 77.0.0.2 200 ?

left-network-1.inet.0: 3 destinations, 3 routes (3 active, 0 holddown, 0 hidden)
<<<<<<<<<<< no routes
left-network-10.inet.0: 3 destinations, 3 routes (3 active, 0 holddown, 0 hidden)

Logs are attached with this bug report

[Nischal Sheth (nsheth)]

Need to generate core using gcore if this happens again.

[OpenContrail Admin (ci-admin-f)]

bug update...