Pipe collisions and missing IPv6 LL addresses
Hi all, I am running bird 1.2.5 as a RS using classical design (one table per peer and one master global). I also added in addition a second table called "collector" that is just populated as the master table with other pipes protocols. Idea is that this table can be polled without touching the master table (never know) 1°) Pipe collisions: -------------------- I have this in my logs: 2011-02-01 10:32:50 <ERR> Pipe collision detected when sending X.Y.Z.T/zz to table collector X.Y.Z.T/zz being of course a valid prefix. Have you guys ever seen this. When does this occur? What is the consequence of a pipe collision? How can I troubleshoot? Note I do have that for IPv4 and IPv6. 2°) Missing LL IPv6 addresses: ------------------------------ I also have some problems with few IPv6 peers which seem not to send me LL addresses. 2011-01-31 16:48:12 <ERR> R655687: Missing link-local next hop address, skipping corresponding routes 2011-01-31 16:48:12 <ERR> R655687: - route xxxx:yyyy::/32 skipped xxxx:yyyy::/32 being a regular IPv6 prefix. The weird thing is that this peering is also flapping... The customer tells me he has the same config with other peerings without any problem. Any hint? Is bird doing more verifications than some vendor routers? Thanks for feedback and sorry if this has already been queried on some forums I'd have missed. -- ------------------------------------------------------------- Jerome Durand Responsable des services aux usagers Services operations & support manager Réseau National de Télécommunications pour la Technologie, l'Enseignement et la Recherche Tel: +33 (0) 1 53 94 20 40 | GIP RENATER Fax: +33 (0) 1 53 94 20 41 | 23-25 rue Daviel E-mail: jdurand@renater.fr | 75013 PARIS http://www.renater.fr | FRANCE --------------------------------------------------------------
Hi Jerome, Am 01.02.2011 11:59, schrieb Jerome Durand:
I have this in my logs:
2011-02-01 10:32:50 <ERR> Pipe collision detected when sending X.Y.Z.T/zz to table collector
we solved the problem by adding an internal community to all routes when they are passed to the master table and rejecting all routes towards collector table that are tagged with that internal community. Regards Bernhard -- Bernhard Hahn DE-CIX Management GmbH e-mail: bernhard.hahn@de-cix.net Lindleystr. 12, 60314 Frankfurt Phone: +49 69 1730 902-34 Geschaeftsfuehrer Harald A. Summa Mobile: +49 171 552 3643 Registergericht AG Koeln, HRB 51135 Fax: +49 69 4056 2716
On Tue, Feb 01, 2011 at 11:59:11AM +0100, Jerome Durand wrote:
Hi all,
I am running bird 1.2.5 as a RS using classical design (one table per peer and one master global). I also added in addition a second table called "collector" that is just populated as the master table with other pipes protocols. Idea is that this table can be polled without touching the master table (never know)
1°) Pipe collisions: --------------------
I have this in my logs:
2011-02-01 10:32:50 <ERR> Pipe collision detected when sending X.Y.Z.T/zz to table collector
Pipes and tables are originally supposed to form a tree, not a dag or even a cycle, you have to set filters in a way that it is a tree from a point of view of each route. You probably have filters on pipe from a local table to the collector table that accepts everything in and nothing out. In that case, if BGP protocol A receives a route, it is added to the local table of proto A, then it is propagated directly to the collector, also to the master table, from that it is propagated to local tables of other BGP protocols and from that again to the collector table (which forms a dag). You need to set filters on a way to the collector table that only routes received directly from a BGP protocols (not routes received through pipe from the master table) are propagated to the collector. BTW, there is not a much difference between polling the master table and the collector table, i don't see a point in that, unless you want do set a different filters.
2°) Missing LL IPv6 addresses: ------------------------------
I also have some problems with few IPv6 peers which seem not to send me LL addresses.
2011-01-31 16:48:12 <ERR> R655687: Missing link-local next hop address, skipping corresponding routes 2011-01-31 16:48:12 <ERR> R655687: - route xxxx:yyyy::/32 skipped
xxxx:yyyy::/32 being a regular IPv6 prefix.
The weird thing is that this peering is also flapping... The customer tells me he has the same config with other peerings without any problem. Any hint? Is bird doing more verifications than some vendor routers?
You need option 'missing lladdr ignore' in BGP protocols. See documentation for details. -- Elen sila lumenn' omentielvo Ondrej 'SanTiago' Zajicek (email: santiago@crfreenet.org) OpenPGP encrypted e-mails preferred (KeyID 0x11DEADC3, wwwkeys.pgp.net) "To err is human -- to blame it on a computer is even more so."
participants (3)
-
Bernhard Hahn -
Jerome Durand -
Ondrej Zajicek