Missing bable routes after configure - bug in bird 2.0.8?

[Nico Schottelius]

Hello,

we are using direct + babel as an IGP and today something "funky"
happened: the routes of one router disappeared and on other routers the
babel entries contained empty router IDs in babel:

bird> show babel entries
babel1:
Prefix                   Router ID               Metric Seqno  Routes Sources
185.203.114.0/23         00:00:00:00:93:4e:c3:fb      0  2204       4       0
195.141.230.102/31       <none>                       -     -       1       0
185.203.112.0/24         <none>                       -     -       2       0
147.78.195.224/27        00:00:00:00:93:4e:c3:fb      0  2204       7       0
193.192.225.72/31        <none>                       -     -       1       0
94.78.2.168/29           00:00:00:00:93:4e:c3:fb      0  2204       4       0
147.78.195.0/29          <none>                       -     -       1       0
185.155.31.0/24          00:00:00:00:93:4e:c3:fb      0  2204       4       0
45.134.132.0/29          00:00:00:00:93:4e:c3:fb      0  2204       4       0
185.155.31.0/26          00:00:00:00:93:4e:c3:fb      0  2204       4       0
2a0a:e5c0:1:f::/64       <none>                       -     -       1       0
2a0a:e5c0:1:e::/64       <none>                       -     -       1       0
2a0a:e5c0:1:d::/64       00:00:00:00:93:4e:c3:fb      0  2204       5       0
2a0a:e5c0:0:9::/64       00:00:00:00:00:00:00:2f     96     1       1       0
2a0a:e5c0:1:8::/64       00:00:00:00:93:4e:c3:fb      0  2204      10       0
2a0a:e5c0:0:6::/64       00:00:00:00:00:00:00:2e     96     2       1       0
2a0a:e5c0:0:5::/64       <none>                       -     -       2       0
2a0a:e5c0:0:4::/64       00:00:00:00:00:00:00:2f     96     1       1       0
2a0a:e5c0:1:4::/64       00:00:00:00:93:4e:c3:fb      0  2204       5       0
2a0a:e5c0:0:3::/64       00:00:00:00:00:00:00:2e     96     2       1       0
2a0a:e5c0:0:2::/64       <none>                       -     -       2       0
2a0a:e5c0:2:2::/64       00:00:00:00:93:4e:c3:fb      0  2204       5       0
2a0a:e5c0:10:1::/64      <none>                       -     -       1       0

After restarting the router, the entries are correct now (router id
00:00:00:00:93:4e:c3:e3):

bird> show babel entries
babel1:
Prefix                   Router ID               Metric Seqno  Routes Sources
185.203.114.0/23         00:00:00:00:93:4e:c3:fb      0  2204       4       0
195.141.230.102/31       <none>                       -     -       1       0
185.203.112.0/24         <none>                       -     -       2       0
147.78.195.224/27        00:00:00:00:93:4e:c3:fb      0  2204       7       0
193.192.225.72/31        <none>                       -     -       1       0
94.78.2.168/29           00:00:00:00:93:4e:c3:fb      0  2204       4       0
147.78.195.0/29          00:00:00:00:93:4e:c3:e3     96     1       1       0
185.155.31.0/24          00:00:00:00:93:4e:c3:fb      0  2204       4       0
45.134.132.0/29          00:00:00:00:93:4e:c3:fb      0  2204       4       0
185.155.31.0/26          00:00:00:00:93:4e:c3:fb      0  2204       4       0
2a0a:e5c0:1:f::/64       00:00:00:00:93:4e:c3:e3     96     1       1       0
2a0a:e5c0:1:e::/64       00:00:00:00:93:4e:c3:e3     96     1       1       0
2a0a:e5c0:1:d::/64       00:00:00:00:93:4e:c3:fb      0  2204       5       0
2a0a:e5c0:0:9::/64       00:00:00:00:00:00:00:2f     96     1       1       0
2a0a:e5c0:1:8::/64       00:00:00:00:93:4e:c3:fb      0  2204      10       0
2a0a:e5c0:0:6::/64       00:00:00:00:00:00:00:2e     96     2       1       0
2a0a:e5c0:0:5::/64       <none>                       -     -       2       0

The "funny" part is that bird 2.0.8 on that router was up and running,
it did see babel neighbors, it did BGP, however only after restarting
bird, the routes were correctly received be neighboring routers.

As you can see in above table there is at least another router that is
affected. I tried `disable babel1` and `enable babel1` on it, however
that did not fix the problem.

We did some amount of `configure` commands on the bird process, but my
understanding was that it should be very similar to restarting it,
without the loss of sessions. The babel protocol *did* previously export
the routes correctly to other routers, but I am 99% sure that it did
stop doing it, until we hard restarted bird.

The babel configuration we use on most of our routers looks like this:

protocol direct {
    ipv4;
    ipv6;
    interface "bond0.*";
}

protocol babel {
        interface "bond0.*" {
                type wired;
        };

        ipv4 {
                export where (source = RTS_DEVICE);
        };
        ipv6 {
                export where (source = RTS_DEVICE);
        };
}

So we are not re-babling like babeld, but only inject local device
routes, as we have 2 redundant routers per network.

Is there a conceptual problem or is this a bug in bird 2.0.8 and in
either way, is there anything we can do to fix it besides restarting
bird on router propagation error?

Best regards,

Nico

p.s.: In case it matters, we are running bird on alpine/musl.

--
Sustainable and modern Infrastructures by ungleich.ch