[quagga-users 14633] Re: Weird BGP Behavoiur

Paul Thornton paul at prt.org
Sun Mar 5 11:53:05 GMT 2017


On 05/03/2017 11:30, Alexis Rosen wrote:
> On Mar 5, 2017, at 4:04 AM, Paul Thornton <paul at prt.org> wrote:
>> It is a royal pain to debug because it seems almost un-reproducible in the lab :(
> Because of this, I suspect the only way to resolve this in a reasonable amount of time is by bisect. Doing that on a production network will suck. But this bug has been running around free for at least a year, and if you're right (below) more than 2.5.
>> FWIW, I have downgraded to across the board as that is the last release of Quagga that I can find that definitely does not have this issue.
> That's an interesting claim, and an important one. All the other reports on this have been about versions 1.x.x, IIRC. Specifically, I was under the impression that did NOT show this bug. If you can reproduce this bug in .24.1 (and .24.0), that will provide new and helpful info on where to look to fix it. How sure are you about this?

Actually, I'm not 100% sure.  I thought I was, but now I'm not!

On our current production routers, we're running - I am not 
certain whether this is just a case of: "that was the version we were 
running pre-upgrade to 1.0 and we know it works so are sticking with it".

Now it just so happens that during some lab staging of a new production 
network for a customer a few weeks ago, I had a FreeBSD box acting as 
"The Internet" which, amongst other things, carried a full routing table.

This *was* indeed running 1.something (I think it was 1.2) and saw the 
problem, reproducibly with v6, and I downgraded that to *I think* as we needed to get testing completed in a bit of a hurry.  I 
cannot check right now what version that was but will be able to clarify 
that next week.  Whatever version I used worked fine.

I may also be in a position where I can use this lab setup briefly after 
our testing (as it is essentially a Quagga router providing transit to a 
'production' network) to do some Quagga testing.  I don't know *what* I 
can do in terms of trying to troubleshoot the code, but I could 
theoretically run something with loads of debug and terrible performance 
in an attempt to track down the issue.  If people can tell me what to 
look for / do here, I can try and fit it in.  Certainly, if nothing 
else, I can use that to absolutely determine which version is the first 
affected one.


Paul Thornton

More information about the Quagga-users mailing list