Repeated announce or withdraw for the same prefix crashes the API

[pwbristow]

You can cause the BGP process to go into the weeds with the following. If you need anything else let me know.

CentOS 6.4 python-2.6.6-37.el6_4.x86_64 bird-1.3.11-1.x86_64 exabgp-3.2.13

Peer Bird config

router id 172.16.112.21;
protocol device { scan time 10; }
protocol direct { interface "eth0"; }
protocol bgp { multihop 1; local as 56673; neighbor 10.0.0.1 as 56673; source address 10.0.0.0; rr client; import all; passive;}

Exacfg

neighbor 10.0.0.0 {
        router-id 10.0.0.1;
        local-address 10.0.0.1;
        local-as 56673;
        peer-as 56673;
        hold-time 180;

        process service-dynamic {
                run /bin/cat /tmp/script;
        }
}

cat /tmp/apiscript
#!/bin/sh
sleep 10
echo "announce route 172.116.10.0/24 next-hop 10.0.0.1 community [123:321]"
sleep 2
echo "announce route 172.116.10.0/24 next-hop 10.0.0.1 community [123:321 123:322]"
sleep 2

#Either of the following lines break things
#echo "withdraw route 172.116.10.0/24 next-hop 10.0.0.1 community [123:321 123:322]"
#echo "withdraw route 172.116.10.0/24 next-hop 10.0.0.1" 
sleep 100

Wed, 02 Oct 2013 07:45:15 | INFO | 64075 | reactor | New Peer neighbor 10.0.0.0 local-ip 10.0.0.1 local-as 56673 peer-as 56673 router-id 10.0.0.1 family-allowed in-open Wed, 02 Oct 2013 07:45:15 | WARNING | 64075 | configuration | Loaded new configuration successfully Wed, 02 Oct 2013 07:45:15 | INFO | 64075 | processes | Forked process service-dynamic Wed, 02 Oct 2013 07:45:16 | INFO | 64075 | network | Connected to peer neighbor 10.0.0.0 local-ip 10.0.0.1 local-as 56673 peer-as 56673 router-id 10.0.0.1 family-allowed in-open (out) Wed, 02 Oct 2013 07:45:25 | INFO | 64075 | processes | Command from process service-dynamic : announce route 172.116.10.0/24 next-hop 10.0.0.1 community [123:321] Wed, 02 Oct 2013 07:45:25 | INFO | 64075 | reactor | Route added to neighbor 10.0.0.0 local-ip 10.0.0.1 local-as 56673 peer-as 56673 router-id 10.0.0.1 family-allowed in-open : 172.116.10.0/24 next-hop 10.0.0.1 community 123:321 Wed, 02 Oct 2013 07:45:26 | INFO | 64075 | reactor | Performing dynamic route update Wed, 02 Oct 2013 07:45:26 | INFO | 64075 | reactor | Updated peers dynamic routes successfully Wed, 02 Oct 2013 07:45:27 | INFO | 64075 | processes | Command from process service-dynamic : announce route 172.116.10.0/24 next-hop 10.0.0.1 community [123:321 123:322] Wed, 02 Oct 2013 07:45:27 | INFO | 64075 | reactor | Route added to neighbor 10.0.0.0 local-ip 10.0.0.1 local-as 56673 peer-as 56673 router-id 10.0.0.1 family-allowed in-open : 172.116.10.0/24 next-hop 10.0.0.1 community [ 123:321 123:322 ] Wed, 02 Oct 2013 07:45:28 | INFO | 64075 | reactor | Performing dynamic route update Wed, 02 Oct 2013 07:45:28 | INFO | 64075 | reactor | Updated peers dynamic routes successfully Wed, 02 Oct 2013 07:45:29 | INFO | 64075 | processes | Command from process service-dynamic : withdraw route 172.116.10.0/24 next-hop 10.0.0.1 Wed, 02 Oct 2013 07:45:29 | INFO | 64075 | reactor | Route found and removed : 172.116.10.0/24 next-hop 10.0.0.1 Wed, 02 Oct 2013 07:45:30 | INFO | 64075 | reactor | Performing dynamic route update Wed, 02 Oct 2013 07:45:30 | INFO | 64075 | reactor | Updated peers dynamic routes successfully Wed, 02 Oct 2013 07:45:30 | ERROR | 64075 | reactor | peer 10.0.0.0 ASN 56673 UNHANDLED PROBLEM, please report Wed, 02 Oct 2013 07:45:30 | ERROR | 64075 | reactor | peer 10.0.0.0 ASN 56673 <type 'exceptions.KeyError'> Wed, 02 Oct 2013 07:45:30 | ERROR | 64075 | reactor | peer 10.0.0.0 ASN 56673 '\x00\x00\x00\x00\x18\xact\n' Wed, 02 Oct 2013 07:45:30 | ERROR | 64075 | | Traceback (most recent call last): Wed, 02 Oct 2013 07:45:30 | ERROR | 64075 | | File "/srv/exabgp-3.2.13/lib/exabgp/reactor/peer.py", line 504, in _run Wed, 02 Oct 2013 07:45:30 | ERROR | 64075 | | for action in self._main(direction): Wed, 02 Oct 2013 07:45:30 | ERROR | 64075 | | File "/srv/exabgp-3.2.13/lib/exabgp/reactor/peer.py", line 471, in _main Wed, 02 Oct 2013 07:45:30 | ERROR | 64075 | | new_routes.next() Wed, 02 Oct 2013 07:45:30 | ERROR | 64075 | | File "/srv/exabgp-3.2.13/lib/exabgp/reactor/protocol.py", line 258, in new_update Wed, 02 Oct 2013 07:45:30 | ERROR | 64075 | | for update in self.neighbor.rib.outgoing.updates(self.neighbor.group_updates): Wed, 02 Oct 2013 07:45:30 | ERROR | 64075 | | File "/srv/exabgp-3.2.13/lib/exabgp/rib/store.py", line 196, in updates Wed, 02 Oct 2013 07:45:30 | ERROR | 64075 | | del dict_nlri[nlri_index] Wed, 02 Oct 2013 07:45:30 | ERROR | 64075 | | KeyError: '\x00\x00\x00\x00\x18\xact\n' Wed, 02 Oct 2013 07:45:30 | ERROR | 64075 | |

[thomas-mangin]

Hi Pete, Could you check if the last commit which fixed another issue did not as a side effect fix this one too as I can not reproduce it.

[tsuna]

I can confirm that exabgp 3.3.0 fixed this issue. I experienced it too on 3.2.13 and the problem went away after upgrading.

[thomas-mangin]

Thank you very much for the feedback, much appreciated.