Comment 5 for bug 1365277

I've tested the proposed patch, and faced a couple of issues:

While configuring the following chain VN1---VNx---VN2 with allow_transit=True on VNx, VN1 routes are re-originated back to the same VN. For exmaple on an adjacent MX router, here is what I see on a VRF sharing VN1's RT:

# run show route table VN1.inet.0
172.16.1.1/32 *[Direct/0] 23:05:20
                                > via lo0.2001
                                [BGP/170] 00:00:00, localpref 100, from 192.168.254.84
                                AS path: 64512 ?, validation-state: unverified
                                > via gr-2/0/10.32769, Push 16

As you can see, VRF loopback address is re-originated in service chain routing instance, and re-annonced back to MX.

Another issue is that as soon as I configure such service policies, vrouter-agent and control processes jump repectively to ~800 ~1200% CPU consumption...