<div dir="ltr"><div class="gmail_extra"><div class="gmail_quote">On Mon, Nov 14, 2016 at 11:55 AM, Ansgar Jazdzewski <span dir="ltr"><<a href="mailto:a.jazdzewski@googlemail.com" target="_blank">a.jazdzewski@googlemail.com</a>></span> wrote:<br><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">Hi,<br>
<br>
i prefer to have only one olsrd running ... do you can point me to the<br>
right configuration to setup olsrd2 in/as Multitopology instance<br></blockquote><div><br></div><div>The problem is that you need a way to feed different metrics into the different topologies...</div><div><br></div><div>Olsrd2 does only come with a single routing metric, so you need a plugin of your own to make use of the multitopology feature (running the same metrics on the interfaces will just give you the same routing table twice).</div><div><br></div><div>There is no "out of the box" way to restrict a topology (or "domain" as I have called it in olsrd2) to certain interfaces.</div><div><br></div><div>Henning</div><div> </div><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
<br>
thanks<br>
<span class="HOEnZb"><font color="#888888">Ansgar<br>
</font></span><div class="HOEnZb"><div class="h5"><br>
2016-11-14 11:09 GMT+01:00 Henning Rogge <<a href="mailto:hrogge@gmail.com">hrogge@gmail.com</a>>:<br>
> While Olsrd2 (the OLSRv2 implementation of OLSR.org) is not Multipath<br>
> capable, it is Multitopology-Capable.<br>
><br>
> So you could have ONE Olsrd2 instance that runs multiple routing metrics,<br>
> creating different routing topologies.<br>
><br>
> You could also run multiple Olsrd2-Instances if you want to split them among<br>
> interfaces (you just have to configure different "locking files").<br>
><br>
> Henning<br>
><br>
> On Mon, Nov 14, 2016 at 11:06 AM, Ansgar Jazdzewski<br>
> <<a href="mailto:a.jazdzewski@googlemail.com">a.jazdzewski@googlemail.com</a>> wrote:<br>
>><br>
>> 2016-11-14 10:37 GMT+01:00 Henning Rogge <<a href="mailto:hrogge@gmail.com">hrogge@gmail.com</a>>:<br>
>> > On Mon, Nov 14, 2016 at 10:17 AM, Ansgar Jazdzewski<br>
>> > <<a href="mailto:a.jazdzewski@googlemail.com">a.jazdzewski@googlemail.com</a>> wrote:<br>
>> >><br>
>> >> Hi,<br>
>> >><br>
>> >> i am try to setup a new smal datacenter and we like to test olsr2 as<br>
>> >> the layer3-dataplane.<br>
>> ><br>
>> ><br>
>> > What kind of multipath are you expecting? Do you just want to use<br>
>> > parallel<br>
>> > ethernet links between certain nodes or do you have "real" disjunct<br>
>> > paths?<br>
>> ><br>
>><br>
>> i think of real multiple path setup i don't like to use bonding or<br>
>> something like it, i have a full mashed datacenter environment the<br>
>> idea is to put vxlan on the top of the routed back-plane, so i can<br>
>> easy setup new networks and all is redundent because of the lower<br>
>> layer3 net: and hopefully easy to scale.<br>
>><br>
>> >><br>
>> >><br>
>> >> wath i have done so far is:<br>
>> >><br>
>> >> bring all my NICs up with a IPv4 link-local and set a host ip on<br>
>> >> loopback.<br>
>> >> olsr is running so far, but i can not see any multipah in my<br>
>> >> routing-table<br>
>> >><br>
>> >> ####<br>
>> >> root@ubuntu:/home/sprd# ip rout show<br>
>> >> default via 172.16.162.1 dev eno2 proto 100 src 172.16.162.167<br>
>> >> metric 2 onlink<br>
>> >> default dev eno2 scope link metric 1003<br>
>> >> <a href="http://169.254.0.0/16" rel="noreferrer" target="_blank">169.254.0.0/16</a> dev eno2 proto kernel scope link src 169.254.8.60<br>
>> >> <a href="http://169.254.0.0/16" rel="noreferrer" target="_blank">169.254.0.0/16</a> dev eno1 proto kernel scope link src 169.254.8.55<br>
>> >> 172.16.162.1 via 172.16.162.1 dev eno2 proto 100 src 172.16.162.167<br>
>> >> metric 2 onlink<br>
>> >> 172.16.162.156 via 169.254.22.61 dev eno2 proto 100 src<br>
>> >> 172.16.162.167 metric 2 onlink<br>
>> >> 172.16.162.161 via 169.254.12.73 dev eno2 proto 100 src<br>
>> >> 172.16.162.167 metric 2 onlink<br>
>> >> 172.16.162.178 via 169.254.9.132 dev eno2 proto 100 src<br>
>> >> 172.16.162.167 metric 2 onlink<br>
>> >> ####<br>
>> >> root@ubuntu:/home/sprd# ip -4 addr show dev lo<br>
>> >> 1: lo: <LOOPBACK,UP,LOWER_UP> mtu 65536 qdisc noqueue state UNKNOWN<br>
>> >> group default qlen 1<br>
>> >> inet <a href="http://127.0.0.1/8" rel="noreferrer" target="_blank">127.0.0.1/8</a> scope host lo<br>
>> >> valid_lft forever preferred_lft forever<br>
>> >> inet <a href="http://172.16.162.168/32" rel="noreferrer" target="_blank">172.16.162.168/32</a> scope global lo<br>
>> >> valid_lft forever preferred_lft forever<br>
>> >> inet <a href="http://172.16.162.167/32" rel="noreferrer" target="_blank">172.16.162.167/32</a> scope global lo<br>
>> >> valid_lft forever preferred_lft forever<br>
>> >> ####<br>
>> >><br>
>> >> so my questions are:<br>
>> >> how can i configure it, or is multipath no possible with olsr2?<br>
>> ><br>
>> ><br>
>> > OLSR and OLSRv2 do not support multipath at the moment (at least the<br>
>> > OLSR.org implementation does not have any extensions for it)... mostly<br>
>> > because the Dijkstra Algorithm gives you only the "fastest" path and it<br>
>> > is<br>
>> > quite difficult (especially for a mesh net environment) to suggest an<br>
>> > algorithm that gives you multiple good paths.<br>
>><br>
>> in my case i have all NICs with the same wight, so i "just" need the<br>
>> second route entry<br>
>> Can i force olsr2 to start the Dijkstra for every leave NIC? and add<br>
>> each NIC into a extra routing Table?<br>
>><br>
>> this sould work in my case? dont know how failover will do, let solve<br>
>> it layter on<br>
>><br>
>> > see <a href="https://tools.ietf.org/html/draft-ietf-manet-olsrv2-multipath-11" rel="noreferrer" target="_blank">https://tools.ietf.org/html/<wbr>draft-ietf-manet-olsrv2-<wbr>multipath-11</a> for<br>
>> > some standardization work on it in the IETF.<br>
>> ><br>
>> > If you have a good idea how to implement an algorithm like this I would<br>
>> > like<br>
>> > to talk about it.<br>
>><br>
>> like i tryed to draw, group outbound-nics and run Dijkstra on each of<br>
>> the links to find a idividual route or the same but from an other<br>
>> source NIC<br>
>> mybe i can try it by running multiple olsr2 deamons?<br>
>><br>
>> > There is also the problem that moving the traffic along these paths is<br>
>> > difficult without a very special setup or source routing (which is<br>
>> > deprecated I think).<br>
>><br>
>> i don't like to have the normal datacenter traffic directly on those<br>
>> links, all Layer2 traffic will be in VxLAN, in the momet all that is a<br>
>> rough idea and i like to build it as open as possible<br>
>><br>
>> thanks for your help,<br>
>> Ansgar<br>
>><br>
>> > Henning Rogge<br>
>> ><br>
>> >><br>
>> >> Tanks<br>
>> >> Ansgar<br>
>> >><br>
>> >> --<br>
>> >> Olsr-users mailing list<br>
>> >> <a href="mailto:Olsr-users@lists.olsr.org">Olsr-users@lists.olsr.org</a><br>
>> >> <a href="https://lists.olsr.org/mailman/listinfo/olsr-users" rel="noreferrer" target="_blank">https://lists.olsr.org/<wbr>mailman/listinfo/olsr-users</a><br>
>> ><br>
>> ><br>
><br>
><br>
</div></div></blockquote></div><br></div></div>