Bird6 high CPU usage

12 Nov 2017

      Hi All,

     I have a loopback IPv6 address on a server. IP addr
= 2dd:1111:1111:1111:1111::1

   I use BGP to announce this to the upstream router. Client traffic comes
in through an interface ens3 connected to the router with destination
address as the loopback. Bird configuration:

   protocol direct {
        description "Local anycast addresses";
        export none;
        interface "lo";
}

# The kernel protocol gives us access to the kernel routing table.
protocol kernel {
        #persist;               # Don't remove routes on shutdown
        scan time 20;           # Scan kernel routing table every 20 seconds
        export all;             # Default is export none
}

# The device protocol is needed in order to see what addresses are local
# to this server.
protocol device {
        scan time 10;           # Scan interfaces every 10 seconds
}

protocol bgp {
        description "Arya";
        import none;
        local as 65001;
        neighbor neighbourIPv6ADDR as 100;
        direct;                 # Neighbors are directly connected
                hold time 240;
        connect retry time 120;
        keepalive time 80;      # defaults to hold time / 3
#export filter packetdns;
                export all;
}

All this works great. Now, when client traffic comes in we get the
following entries in the routing table:

.
.
.
CLIENTIP::136 via 2dd:1111:1111:1111:1111::1 dev ens3  metric 0
    cache
CLIENTIP::5 via 2dd:1111:1111:1111:1111::1 dev ens3  metric 0
    cache
CLIENTIP::13 via 2dd:1111:1111:1111:1111::1 dev ens3  metric 0
    cache
CLIENTIP::15 via 2dd:1111:1111:1111:1111::1 dev ens3  metric 0
    cache
.
.
.

This table grows to  ip -6 r s |wc -l
8296

And changes constantly.

Now, the top output shows bird using a lot of CPU:
2205 root      20   0    6924    780    560 R  26.2  0.0   0:06.14 bird6

Since my configuration will never make use of these entries, is there a way
to disable this scanning behaviour in bird?

Thanks!
-- 
Saksham Manchanda