>
>
> bgpd: Quagga, BGP feature Best External
>>
>> http://git.savannah.gnu.org/cgit/quagga.git/commit/?h=refs/heads/volatile/patch-tracking/4/proposed/ff&id=d30eee229f5f85205753c95c934993a7f8644176
>> Needs formatting fixes and review
>>
>
> Needs formatting fixes at a minimum (submitter could do those), plus some
> review as it's touching pretty core and critical BGP functionality (needs
> others).
>
>
I wanted to pass on that I did some preliminary testing on these diffs and
found one significant issue. It could be related to my port of the changes
to our code base, but wanted to pass on the results in case the submitter
would like to test his changes to see if the problem also exists there.
Test environment:
A-------B----
\ /
\ /
C----
A, B, and C are iBGP peers
B and C are receiving the same prefix via eBGP
B has local pref set so it is the preferred path out of the iBGP network
With C running the code with best external support, it receives and prefers
the path through B, but also advertises his best external prefix to A.
This works fine. At this point, A contains both the path through B and the
path through C, with the path through B being preferred.
The problem I encountered occurred when the prefix through B is withdrawn
and the path through C becomes best. At this point, A has the path through
C (as it should), but when the route is re-learned on B and re-advertised
into iBGP with higher local pref, C receives this prefix, installs it as
best and withdraws it's own. At the end of this sequence, A only contains
the path through B.
Again, this could be because I botched the port of the diffs, but thought
the results of the preliminary testing worthy of sharing. If not, sorry
for the noise.
--
Don Slice
Cumulus Networks
_______________________________________________
Quagga-dev mailing list
[email protected]
https://lists.quagga.net/mailman/listinfo/quagga-dev
