Full BGP table in VRF

RouterOS Forwarding Protocols
1

Can CCR can support full BGP table in VRF?

I have an ISP setup with MPLS and BGP-free core. Route Reflectors (RRs) are in place. Internet traffic is in a VRF between the transit routers and CE routers. Two Transit routers (TR) peer with different upstream providers and each have full BGP tables. iBGP for MPLS configured as such: TRs<—>RRs<—>CEs

All routers running 6.40.4
Transit, RR, and P routers are all CCR1072
CE routers are CCR1016

IGP is OSPF
Static and Connected BGP prefixes are exchanged correctly between transit and CE routers.

/routing bgp instance vrf
add redistribute-connected=yes redistribute-static=yes routing-mark=inet

I then make a change to a transit router:
/routing bgp instance vrf
add redistribute-connected=yes redistribute-static=yes redistribute-other-bgp=yes routing-mark=inet

At this point, Internet BGP prefixes begin to propagate to the CE routers as expected. However, after less than a minute the transit router crashes and reboots. The transit router will not stablize until this line is reverted back.

2

Hi,
Any particular reason for doing it like this and not just using the global (main) table?

Nevertheless, I’ve tried in labs things like this, although not using CCR but CHR (using 2 xeon processors, 8 GB of ram) and even though they take the full table the first time with no issue, when I tested this, I did see instabilities and some locks on the routing process, although I have to admit it’s been more than a year since I tried it perhaps this is not the case anymore.