Hey Ondrej, On Tue, 11 Sep 2018 at 14:01, Ondrej Zajicek <santi...@crfreenet.org> wrote:
> You are right, there is a lock to avoid run two instances for the same > neighbor. Although the primary reason for this lock is to have proper > ordering of protocol startups during reconfiguration. We should fix it > to avoid this issue. Many thanks!. And thank you for your work on bird. I spent maybe 50h trying to do this on IXIA, but just does not scale. On bird I could do it in about 10h, with Bird using just 25GB of memory. > It should, because incoming connection has random/ephemeral source port, > so it is not verified. You're right, I noticed that after posting, that I can't check anything, so nothing to fix there, but didn't want to add noise :) -- ++ytti
