- Bird-users - Mailing Lists

Setting BGP metric based on OSPF cost to peer
by Lukas Haase 18 Jun '26

18 Jun '26

Hello, I have a network with 2 bird routers and one Mikrotik RouterOS router; internal routing via OSPF. The bird boxes announce a /24 prefix via eBGP and then there are 3 iBGP sessions: bird1<-->bird2, bird1<-->mikrotik and bird2<-->mikrotik. For now, both bird1 and bird2 send Mikrotik a default route (imported via eBGP; might be full table later on). Mikrotik should then make the default route active whose gateway has the shortest path via OSPF metric. To my understanding, this is "best practice" and the right way to do it: OSPF defines the best route within my network and on top, iBGP defines the best route to the internet (via the internal network). If my understanding is correct, when Mikrotik receives the two default routes and all BGP metrics are identical it should eventually resort to the underlying IGP metric (i.e., lowest OSPF cost to the gateway of the default route) as a tie breaker. (Please correct me if this is not the common case). Turns out Mikrotik is totally broken here and ignores IGP metric (OSPF cost) completely, resulting in sub-optimal installation of the default route. I could overwrite this STATICALLY (e.g. by setting bgp-igp-metric in a route filter) but the point is that the optimal path is defined DYNAMICALLY via the underlying IGP (OSPF). Super frustrating. Now my idea is if I could turn this around and let the birds prioritize the routing: Is it possible to attach a BGP metric on export that equals the OSPF cost of the BGP peer (i.e., Mikrotik)? How would I do this and is it a good idea? And if so, can it be done such that an update is sent anytime the underlying OSPF cost changes? Thanks, Luke

1 0

BIRD 3.3.0 crashes on "birdc show route for 209.24.0.0/24"
by Robert Scheck 16 Jun '26

16 Jun '26

Hello all, unfortunately, BIRD 3.3.0 as shipped by EPEL 9 crashes at the following: $ birdc show route for 209.24.0.0/24 BIRD 3.3.0 ready. Connection closed by server $ Jun 9 18:34:47 kernel: bird[17874]: segfault at 0 ip 0000562a987ed574 sp 00007ffed8687a30 error 4 in bird[562a987a1000+11d000] likely on CPU 18 (core 2, socket 0) Jun 9 18:34:52 systemd-coredump[17976]: Process 17874 (bird) of user 990 dumped core.#012#012Stack trace of thread 17874:#012#0 0x0000562a987ed574 rt_show_net (/usr/sbin/bird + 0x61574)#012#012Stack trace of thread 17876:#012#0 0x00007f18d1f032cf n/a (n/a + 0x0) #012ELF object binary architecture: AMD x86-64 Jun 9 18:34:52 systemd[1]: bird.service: Main process exited, code=dumped, status=11/SEGV The correct answer of above command should be "Network not found", which it is on at least BIRD 2.16. Btw, "birdc show route 209.24.0.0/24 filtered" crashes the same way on BIRD 3.3.0. The correct answer would also be "Network not found". But as "birdc show route for 1.1.1.1" and "birdc show route for 1.1.1.0/24" both work fine, this only seems to affect non-existing routes? Do you need anything else for debugging? Regards, Robert

4 3

Re: Releases on bird.network.cz no longer updated
by Leon Grasmeijer 15 Jun '26

15 Jun '26

On 13/06/2026 13:48, Ondrej Filip wrote: > On 11/06/2026 09:33, Leon Grasmeijer wrote: >> Hi there, >> >> Are https://bird.network.cz/?download and >> ftp://bird.network.cz/pub/bird/ no longer being updated? >> I see there are some versions missing that have been announced and >> are present on https://bird.nic.cz/get-bird/. > > Hi! > Yes, we are migrating everything to bird.nic.cz. Do you still see a > need for FTP? Hi Ondrej, thank you for your response. I'll make sure that we'll pull from bird.nic.cz from now then. I don't see a need to keep using FTP. > > Ondrej > > >> >> Kind regards, >> Leon >

1 0

Re: Releases on bird.network.cz no longer updated
by babydr DBA James W. Laferriere 15 Jun '26

15 Jun '26

Hello Ondrej , On Sat, 13 Jun 2026, Ondrej Filip wrote: > On 11/06/2026 09:33, Leon Grasmeijer wrote: >> Hi there, >> >> Are https://bird.network.cz/?download and ftp://bird.network.cz/pub/bird/ >> no longer being updated? >> I see there are some versions missing that have been announced and are >> present on https://bird.nic.cz/get-bird/. > > Hi! > Yes, we are migrating everything to bird.nic.cz. Do you still see a need for > FTP? > > Ondrej While I admint to have not looked very hard . rsync might be a decent addition ! ... Tia , JimL -- +---------------------------------------------------------------------+ | James W. Laferriere | System Techniques | Give me VMS | | Network & System Engineer | 3237 Holden Road | Give me Linux | | jiml(a)system-techniques.com | Fairbanks, AK. 99709 | only on AXP | +---------------------------------------------------------------------+

2 1

Re: CentOS 7 repo not work
by Robert Scheck 13 Jun '26

13 Jun '26

On Sat, 13 Jun 2026, zx--- via Bird-users wrote: > Some of my servers are still using CentOS 7 and I'm following https://bird.nic.cz/get-bird/ to install bird, but it looks like CentOS 7 repo is broken? CentOS Linux 7 is EOL since 2024-06-30, thus nearly for two years. And personally, I highly appreciate that this repository is no longer working (sorry) as it really urges admins to get rid of their legacy systems. https://phoenixnap.com/kb/migrate-centos-to-rocky-linux and others provide steps how to inline upgrade existing CentOS Linux 7 systems, if it helps. Regards, Robert

1 0

None
by None 13 Jun '26

13 Jun '26

not be turned off in iBGP configurations, but this is also true on=20 othe platforms (cisco, juniper, ...). On the other hand having gateway direct (and thus multihop turned off) = might=20 be needed when using with sorted tables. > >=20 > > Removing 'gateway direct' fixes the error (but obviously not the > > behaviour). > >=20 > > This is running in an network container if that's of any interest. > >=20 > > Any ideas? > >=20 > > (apologies for a potential dupe - I sent this to what I think must = have > > been an old list email address) --=20 SP5474-RIPE Sergey Popovich

1 0

None
by None 13 Jun '26

13 Jun '26

bgp_community.add( [] ) bgp_ext_community.add( [] ) should treat arguments like (e)clist. It would be more logical. > But it should report error during filter: > bird: filters, line 169: Can't add set It does not reports anything. Moreover, if change [] brackets with () brackets: cmn = ( (ro,MyASN,10856), (ro,MyASN,2) ); ... there is no syntax error, but it should be? -- Alexander Shikov Technical Staff, Digital Telecom IX Tel.: +380 44 201 14 07 http://dtel-ix.net/

1 0

None
by None 13 Jun '26

13 Jun '26

ecurity=20 reasons): OSPF: Bad packet from 10.1.2.3 - authentication failed As the LOG messages increases, we try to debug this and enable=20 debug protocols { packets } so we see this as the case of the LOG messages: 2014-03-07 10:59:20 <TRACE> aclctrans: OSPF_auth: lower sequence number= (rcv=20 13229786, old 13229787) 2014-03-07 10:59:20 <ERR> OSPF: Bad packet from 10.1.2.3 - authenticati= on=20 failed 2014-03-07 10:59:20 <TRACE> aclctrans: OSPF_auth: lower sequence number= (rcv=20 13229786, old 13229787) 2014-03-07 10:59:20 <ERR> OSPF: Bad packet from 10.1.2.3 - authenticati= on=20 failed 2014-03-07 10:59:20 <TRACE> aclctrans: OSPF_auth: lower sequence number= (rcv=20 13229786, old 13229787) 2014-03-07 10:59:20 <ERR> OSPF: Bad packet from 10.1.2.3 - authenticati= on=20 failed 2014-03-07 10:59:20 <TRACE> aclctrans: OSPF_auth: lower sequence number= (rcv=20 13229786, old 13229787) 2014-03-07 10:59:20 <ERR> OSPF: Bad packet from 10.1.2.3 - authenticati= on=20 failed 2014-03-07 10:59:20 <TRACE> aclctrans: OSPF_auth: lower sequence number= (rcv=20 13229786, old 13229787) 2014-03-07 10:59:20 <ERR> OSPF: Bad packet from 10.1.2.3 - authenticati= on=20 failed 2014-03-07 10:59:20 <TRACE> aclctrans: OSPF_auth: lower sequence number= (rcv=20 13229786, old 13229787) 2014-03-07 10:59:20 <ERR> OSPF: Bad packet from 10.1.2.3 - authenticati= on=20 failed We do not have any clue about where the decremented sequence number cam= e from. As this is an internal network and we have an helo interval of 5 second= we can=20 not imagin that this could be an "late" or "old" paket. At the Moment we use mainly bird 1.3.11 with some exceptions where we h= ave=20 bird 1.4.0. The problem seems not only bird generated as we also have f= rom IPs=20 which are router in our network. --=20 Mit freundlichen Gr=FC=DFen --------------------------------- Daniel Wendler

1 0

None
by None 13 Jun '26

13 Jun '26

as routing daemon using single socket 4 core Intel Xeon X3470 (2.93 Ghz= ), with 4GB RAM and 4 multiqueue NIC (like Intel 82576, 82580, ...) routes= 1,5 - 1,8 Gbit/sec at PPS ~300 000 with ~ 98% of CPU time spent at the lowest= frequency (1.2 Ghz). Software part is custom Linux distribution based o= n Debian GNU/Linux 7.x (wheezy), and custom Linux kernel based on 3.2 sta= ble=20 branch, build for x86_64 architecture. So Intel Xeon X3470 is much likely very powerful for such setups (pure forwarding with no traffic policy). On other deployments I have similar results: CPU: Intel Xeon X3470, 2.93 RAM: 2GB-6GB Task: Dual stack IPv4/IPv6, shaping with HTB (U32 hash classifier), 802.1ad (QinQ) to customer end (this turns off hardware offloading task= s on NICs), uRPF (Reverse Path filtering from customer side to prevent IP= spooofing) for both IPv4/IPv6, few iptables rules (connection tracking = is turned off) and aggressive ipset usage (billing hooks, DDoS, etc). Traffic (In:Out): 2Gbit/sec:1.5Gbit/sec, PPS <450 000. Peak load: <80% on all cores. Also having threads (Hyper-Threading) has no good effect on traffic for= warding workloads, as CPU cache is shared between physical cores, causing more = cache misses and thus lower performance. Amount of RAM seems not wery important, hardware with starting from 2GB= RAM=20 sizes is enough to store routing cache information. >=20 > Thanks to the dev team, and kind regards. >=20 > Tigran Zakoyan. --=20 SP5474-RIPE Sergey Popovich

1 0

None
by None 13 Jun '26

13 Jun '26

A iperf seemed to pick a single eth0 or eth1 but not both interfaces and use it to send out traffic on all 4 difference processes. Ideally how would one allow a single application (in this case iperf) with multiple threads/processes to send 2Gbit worth of traffic? Ideally to a logical interface and it's hash automatically to two difference eth0 and eth1? I've looked at setting (krt_prefsrc) to source the address as the dummy0 interface on each host. Would that be the answer? thanks, -bn 0216331C --047d7b6d7d5a9e0f7a04dc1709ba Content-Type: text/html; charset=UTF-8 Content-Transfer-Encoding: quoted-printable <div dir=3D"ltr">I'm trying out OSPF ECMP on BIRD (with ECMP set and en= abled on kernel) and on hosts A and host B are configured as below and conn= ected via a switch, the dummy0 interface is used as the loopback address fo= r two physical nics.<div> <br></div><div style>host A</div><div style>eth1 - /30 - OSPF</div><div sty= le>eth0 - /30 - OSPF</div><div style>dummy0 - /32</div><div style><br></div= ><div style><div>host B</div><div>eth1 - /30 - OSPF</div><div>eth0 - /30 - = OSPF</div> <div>dummy0 - /32</div></div><div><br></div><div>From host A I do "ipe= rf -c x.x.x.x -P 4" to host B's dummy0 interface, host A iperf see= med to pick a single eth0 or eth1 but not both interfaces and use it to sen= d out traffic on all 4 difference processes. Ideally how would one allow a = single application (in this case iperf) with multiple threads/processes to = send 2Gbit worth of traffic? Ideally to a logical interface and it's ha= sh automatically to two difference eth0 and eth1?</div> <div><br></div><div>I've looked at setting (<span style=3D"line-height:= 19px;text-align:justify">krt_prefsrc) to source the address as the dummy0 i= nterface on each host. Would that be the answer?</span></div><div><br><div> thanks,</div><div><div>-bn<br>0216331C</div> </div></div></div> --047d7b6d7d5a9e0f7a04dc1709ba--

1 0