LLGR : control node sending EOR early before converging with agent

[venu kolli]

Build : 3.2.0.0-2744

Topology mentioned below with control(host1) also peered with MX .

GR is configured on MX and control nodes 1 and 2.

kill control node process , Traffic from MX to VM goes on , as it went to GR state.

bring back control node , Traffic stops for a 5- 10 sec before it resumes .

Observed that control node is sending EOR , before it could update the routes due to which mx is withdrawing the prefixes due to which traffic is being dropped.

env.roledefs = {
    'all': [host1, host2, host3],
    'cfgm': [host1],
    'openstack': [host1],
    'control': [host1, host2],
    'compute': [host3],
    'collector': [host1],
    'webui': [host1],
    'build': [host_build],
    'database': [host1],
}

[Nischal Sheth (nsheth)]

Fixed in 3.2 and master via the following reviews:

https://review.opencontrail.org/#/c/25566/
https://review.opencontrail.org/#/c/25565/