bgp soft reconfiguration inbound

Wolfgang Hennerbichler wh at univie.ac.at
Tue Dec 1 13:21:54 CET 2009


On Dec 1, 2009, at 13:09 , Ondrej Zajicek wrote:

> On Tue, Dec 01, 2009 at 10:48:56AM +0100, Wolfgang Hennerbichler wrote:
>>> Every route passes through at most one pipe according to your expected
>>> filter behavior?
>> 
>> well, a route can be distributed from one table to let's say 10 other
>> tables by 10 pipes. you remember we do things a little different. but
>> the pipes are set like above. so there can't be a triangle or something
>> that sends routes from table A, to table B, to table C and back to A.
>> That shouldn't be possible.
> 
> My question is whether it should be possible taht route was sent from
> table A through table B to table C (where A != C). According to what
> you wrote about pipe filters it shouldn't be possible, unless you have
> someting like two tables associated with one IP address.

no, it should actually not be possible. I'll look into that deeper though. 

> 
>>>> Do the pipes ignore the filters at any time? 
>>> 
>>> I think filters are not ignored.
>> 
>> ok. I think it's the loop prevention thing that's in place before the
>> filter which causes this, this is very certain. I also checked quite
>> some routes where bird thinks they're looping, and all of them are where
>> they should be. Are there chances you'll fix the loop checking code so
>> that it at least checks always after filtering?
> 
> Your problem is probably related to the problem mentioned by me (loop check
> before filter) but according to my understanding of the code, it shouldn't
> happen in your case. It might happen only when route was successfully
> sent from A through B to C (or through a longer chain).

as stated above, shouldn't happen. 

> Could you enable logging of routes ('debug protocols { routes }') and
> send me at least appropriate part (messages related to route triggered
> loop check) of a log?

ok. I'm in the process of setting up a test-server right now to figure things out. 

> 
>>>> PS: We've got BIRD running at VIX in Beta now, about 12 participants,
>>>> no crashes, no problems at all (except for the loop notices)
> 
> BTW, we have implemented route refresh, it will be in the next release.

great news, thanks! 

> But route refresh solves just changes in import filters.

good. 

> Do you also need similar kind of 'soft reconfiguration' for export
> filters?

I don't need it currently, but if it's easy to implement it would make bird a little more feature rich, which will never be a bad idea, and of course it's nice to have that feature - I see me asking for it in ~6 months, so if you can easily do it now, go for it :) 

Wolfgang

> -- 
> Elen sila lumenn' omentielvo
> 
> Ondrej 'SanTiago' Zajicek (email: santiago at 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."

-- 
www.vix.at | www.aco.net
wh at univie.ac.at | WH844-RIPE
Vienna University Computer Center



More information about the Bird-users mailing list