[Olsr-users] Dynamic / Manual route selection in case of bad route

[Henning Rogge]

On Freitag 27 Februar 2009 09:50:44 Michael Rack wrote:
> Hi List,
>
> i use OLSR to announce my public switched IP-Adresses over 3
> DSL-Connections with different connection-speeds and different
> connection-types.
>
> * First DSL-Connection is wireless connected by 2048 / 2048 kBit.
> * Second DSL-Connection is wired SDSL connected by 2048 / 2048 kBit.
> * Third DSL-Connection is wired ADSL connected by 4096 / 1024 kBit.
>
> I use OLSR to manage my up- und download-directions with the WEIGHT-Flag
> in the olsrd configuration-file.
Which version of OLSR are you using ?

> Here is my Problem: My first DSL-Connection sometimes have packetloss or
> results in a high roundtriptime. The clients behind my Router can access
> the internet but have a bad connection.
>
> I like to deselect my first DSL-Connection in OLSR to stop routing
> through this device without downtime...
>
> > Interface "bond0" "bond1" "bond2"
> > {
> >   HelloInterval       2.0
> >   HelloValidityTime   6.0
> >
> >   TcInterval          2.0
> >   TcValidityTime      6.0
> >
> >   MidInterval         2.0
> >   MidValidityTime     6.0
> >
> >   HnaInterval         2.0
> >   HnaValidityTime     6.0
> > }
> >
> > Interface "bond0"
> > {
> >   Weight          1
> > }
> >
> > Interface "bond1"
> > {
> >   Weight          3
> > }
> >
> > Interface "bond2"
> > {
> >   Weight          5
> > }
>
> My currently solution is to use netfilter and add a rule to drop packets
> on bond0 arrives / sent on port 698. After 6 seconds, my route-table
> changes as expected.
>
> > iptables -A INPUT -i bond0 -p udp --dport 698 -j DROP
> > iptables -A OUTPUT -o bond0 -p udp --dport 698 -j DROP
>
> But this is a bad solution?!?
so you want to block an OLSR instance from the network without 
killing/restarting the OLSR ? Yes, IpTables might be a good sollution for 
this.

> Is there a other way to deselect an interface for routing-reasons,
> without any downtime?
The problem is that OLSR handles only "OLSR quality". An external traffic will 
not be measured by OLSR itself because HNA Interfaces have no concept of 
"quality" (they just belong to the node which announce the HNA).

> OLSR should check the the interface quality (missing hello-messages,
> triptime of hello-messages) and then select a better interface that have
> the best quality.
>
> I don't like to be a fulltime admin for the manual network-selection.
>
> The B.A.T.M.A.N Project selects the perfect interface by collecting
> informations about the interface quality (missing packages, packet trip
> time and so on).
>
> In final i need "weight" to prefer bond0 for upstream. On the other side
> of the VPN i prefer bond2 for upstream. So i can get 4096kBit download
> over ADSL and 2048kBit upload over Wireless.
I don't think this is possible, especially if you use public-IP space. How 
should the internet should know which interface has the better uplink to your 
network ?

Henning

-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 198 bytes
Desc: This is a digitally signed message part.
URL: <http://lists.olsr.org/pipermail/olsr-users/attachments/20090228/27199cb6/attachment.sig>