[Olsr-dev] please test 0.5.6 prior to -r4 release

Harald Geyer (spam-protected)
Wed Feb 4 06:25:09 CET 2009


 
> Am Tuesday 03 February 2009 11:36:30 schrieb L. Aaron Kaplan:
> > yes, like last time:
> >
> >    --verbose please.
> >
> > Can you describe how to repeat in a controlled manner?
> > Then it often is just a matter of a day to get it fixed.

Sorry, for not including this in my last mail, but reproducing this is
rather trivial: 

1) Start the new olsrd on an interface that can reach the funkfeuer network.
2) wget -qO - http://localhost:2006/topo

Does your question imply that you can't reproduce this bug on your
side. That would be ... interessting.
 
> I binary dump with "-w <filename> -s 0" might be even better I think.

Ok, tried this now in a more controlled experiment:
1) start tcpdump
2) start olsrd
3) take a dump of the topology every 30 seconds for about 5 Minutes

It wasn't trivial to get a dump of this size from a fonera, but here
is what I got:
http://texas.funkfeuer.at/~harald/olsrd-tip/olsr.dump.gz
http://texas.funkfeuer.at/~harald/olsrd-tip/topo.tgz

After looking at this data there is a clear pattern: At first almost all
entries are marked infinite, but the number decreases over time...
(spam-protected):~/0xff/topo.log/real$ grep -c INF topo.2*
topo.2009-02-04T05:50:15+0100:58
topo.2009-02-04T05:50:45+0100:151
topo.2009-02-04T05:51:19+0100:45
topo.2009-02-04T05:51:54+0100:28
topo.2009-02-04T05:52:29+0100:30
topo.2009-02-04T05:53:04+0100:34
topo.2009-02-04T05:53:39+0100:23
topo.2009-02-04T05:54:15+0100:18
topo.2009-02-04T05:54:49+0100:26
topo.2009-02-04T05:55:25+0100:22

Does olsrd mark edges, that are disconnected from the graph, as infinite?
Actually, that would be a cool (but a bit obscure) debugging feature.
If only people knew that ...

HTH,
Harald




More information about the Olsr-dev mailing list