Routing

Dear all ,
While trying to migrate from my current two bgp sessions with 800mbps each to a third one with 1.8gbps capacity founded the following behaviour.

I began moving my pools one by one from my old peers to the new one , deleting the routes adveritse from the old ones and adding it to the new one.

Everything was ok and the traffic was growing according I added more routes to new peer until I moved the fourth and last pool.

After commited , all the traffic beging coming from the old peers eventhough I was not advertise routes over them and the traffic from the new peer falled to a minimun.

Very strange...

So ...

Should I shutdown the old sessions ? (I would like to keep them just in case)

Should I add some metric preference parameter so my provider sends all the traffic over this new session?

Following my three sessions config.

leandro@mdz-gc-he-mx960# show protocols bgp   
traceoptions {
    file bgp size 10m files 10 world-readable;
    flag state detail;
    flag policy;
    flag route;
    flag update;
}
mtu-discovery;
log-updown;
group first_bgp {
    type external;
    local-address 1.1.1.1;
    import RCV_FROM_PROVIDER;
    family inet {
        unicast {
            prefix-limit {
                maximum 500000;
            }
            rib-group to_CORE;
        }
    }
    export ADVERTISE_1;
    peer-as 4444;
    multipath;
    neighbor 1.1.1.2 {
        description first_bgp;
        multihop;
        local-as 5555 loops 2;
    }
}
group second_bgp {
    type external;
    local-address 2.2.2.1;      
    import RCV_FROM_PROVIDER;             
    family inet {                       
        unicast {                       
            prefix-limit {              
                maximum 500000;         
            }                           
            rib-group to_CORE;          
        }                               
    }                                   
    export ADVERTISE_2;             
    peer-as 4444;                       
    multipath;                          
    neighbor 2.2.2.2 {          
        multihop;                       
        local-as 5555 loops 2;         
    }                                   
}                                       
group third_bgp {                   
    type external;                      
    local-address 3.3.3.1;       
    import RCV_FROM_PROVIDER;             
    family inet {                       
        unicast {                       
            prefix-limit {              
                maximum 500000;         
            }                           
            rib-group to_CORE;          
        }                               
    }                                   
    export ADVERTISE_3;             
    peer-as 5555;                       
    multipath;                          
    neighbor 3.3.3.3 {           
        description "third_bgp";
        multihop;                       
        local-as 5555 loops 2;         
    }                                   
}        

Any sugestion would be wellcome.

Leandro.

I would create an import policy for traffic coming in from the old peers to change the "local preference" value to 40 or any value lower than 100, so they will be less desirable and not selected, but will be available in he event that one of the new ones fail. Ten you can decommission the old peer laer when you are comfortable.

If it is the problem with incoming traffic then you can put an export policy to change MED (metric) to higher value for old peer and to lower value for new peers. Or else you can prepend AS-PATH multiple times, if your SP doesn't support MED.

If the problem is with outgoing traffic then you can put an import policy as mentioned above to change local preference accordingly.