Constant delete / add route after upgrade to 1.6.3

[Maria Jan Matějka]

On May 8, 2018 3:44:20 PM GMT+02:00, Xavier Trilla <xavier.trilla at clouding.io> wrote:
> Hi,
> 
> Some extra Info: I've changed the scan time from 20 to 60 on the
> kernel and the CPU load on the machine has reduced, but I still see
> too many route changes.
> 
> Thanks!
> 
> 
> Saludos Cordiales,
> Xavier Trilla P.
> Clouding.io<https://clouding.io/>
> 
> ¿Un Servidor Cloud con SSDs, redundado
> y disponible en menos de 30 segundos?
> 
> ¡Pruébalo ahora en Clouding.io<https://clouding.io/>!
> 
> De: Bird-users <bird-users-bounces at network.cz> En nombre de Xavier
> Trilla
> Enviado el: martes, 8 de mayo de 2018 14:02
> Para: bird-users at network.cz
> Asunto: Constant delete / add route after upgrade to 1.6.3
> 
> Hi,
> 
> We've upgraded recently to 1.6.3 (We were using 1.2.5, which we had
> running for like 8 years!) and after the upgraded we are seeing a
> weird behavior.
> 
> When running ip monitor route in the machine, looks like bird is
> constantly removing and adding routes, which could be considered
> normal, but the thing is that it deletes a route and then adds the
> same route again (Those routes are received via BGP):
> 
> Heres an example:
> 
> 103.72.2.0/24 via 91.126.95.153 dev vlan111  proto bird
> Deleted 103.72.1.0/24 via 91.126.95.153 dev vlan111  proto bird
> 103.72.1.0/24 via 91.126.95.153 dev vlan111  proto bird
> Deleted 177.75.72.0/24 via 91.126.95.153 dev vlan111  proto bird
> 177.75.72.0/24 via 91.126.95.153 dev vlan111  proto bird
> Deleted 185.165.122.0/24 via 91.126.95.153 dev vlan111  proto bird
> 185.165.122.0/24 via 91.126.95.153 dev vlan111  proto bird
> Deleted 196.46.242.0/24 via 91.126.95.153 dev vlan111  proto bird
> 196.46.242.0/24 via 91.126.95.153 dev vlan111  proto bird
> Deleted 64.70.30.0/24 via 91.126.95.153 dev vlan111  proto bird
> 64.70.30.0/24 via 212.80.191.185 dev vlan2  proto bird
> 
> At the same time we are seeing the following messages on bird log:
> 
> May  8 13:56:39 icewall-01 bird: Kernel dropped some netlink messages,
> will resync on next scan.
> May  8 13:57:02 icewall-01 bird: Kernel dropped some netlink messages,
> will resync on next scan.
> May  8 13:57:27 icewall-01 bird: Kernel dropped some netlink messages,
> will resync on next scan.
> May  8 13:57:32 icewall-01 bird: Kernel dropped some netlink messages,
> will resync on next scan.
> May  8 13:57:59 icewall-01 bird: Kernel dropped some netlink messages,
> will resync on next scan.
> 
> Any ideas why this might be happening?
> 
> Thanks!
> 
> Saludos Cordiales,
> Xavier Trilla P.
> Clouding.io<https://clouding.io/>
> 
> ¿Un Servidor Cloud con SSDs, redundado
> y disponible en menos de 30 segundos?
> 
> ¡Pruébalo ahora en Clouding.io<https://clouding.io/>!

Hello, have you tried to switch on BGP protocol tracing? Aren't the routes really sent in this way? Haven't you seen some protocol going up and down repeatedly or something like that?

This all is to find out whether the routes are flapping on the input side or in the kernel binding in BIRD.

The error message is just a warning that the buffer between kernel and BIRD has overflown which may be due to high cpu load -- but it has to be due to something.

Could you please also send your config to see whether it can be seen from it directly?

Thanks!
Maria