[OLSR-users] Anyone tried two nodes with internet connection?

Sven-Ola Tuecke (spam-protected)
Wed Jan 10 14:22:57 CET 2007


Andreas, et.al.,

the [ping,traceroute] test needs to be carefully designed. For this reason, 
freifunk firmware uses policy routing and a very downstripped version of 
dyngw named "dyngwplain" (no ping test in the plugin itself, it announces 
HNA if default route is in the system standard routing table) as well as a 
cronjob for this:

- traceroute to root nameservers every minute (limited by N hops)
- if fail, remove standard default route (hence remove hna4)
- add old default route to policy routing table "only this device"
- continue tracerouting (over the local-iface-defroute)
- if connection comes up, restore def route which restores HNA

One drawback: During outage, the device itself has no connectivity via the 
OLSR fallback default route. Needs more policy routing tricks if that should 
be the case. But of course the [win|mac|lin] PC behind that box will 
function via both defroutes...

Using busybox traceroute has the advantage, that a gateway needs to 
transport UDP in order to succeed the test. UDP is blocked often due to 
braindead firewalls / users / settings. I do not call a TCP-only gw an 
internet gateway - I personally would never accept if any telco blocks this 
kind of signalling. Of course ping/icmp is an alternative test (if no UDP), 
but you need reliable ping peers for that.

// Sven-Ola

"Andreas "Tønnesen"" <(spam-protected)> schrieb im Newsbeitrag 
news:(spam-protected)
>
> Ok, it's been a long time since I've looked at the dyn-gw plugin so I
> might be wrong here. But isn't the whole point with the pinging to
> determine if the Internet connection has gone bad? Now if you remove the
> route and the pings go via the OLSR network then the node will belive that
> it has an Internet connection and therefore will start announcing it
> again. AFAIK the GWs should be dedicated GWs that will loose Internet
> connectivity if their uplink goes down. They should not be boxes that
> depend on Internet connectivity via the OLSR network.
>
> So IMO the current behavior is the "correct" one as long as the ping is
> routed normally by the IP-stack. I guess one alternative could be to set
> up host routes to the pinged hosts...
>
> - Andreas
>
>> More things to add. I am trying:
>> Pings to Google.com from Node1 -- This works with Internet A up and
>> running.
>> Pings to Google.com from Node1 -- FAILS when I remove the link to
>> InternetA.
>>
>> I would assume it should hop via Node2 and hit google.com via InternetB.
>>
>> Problem is that since Node1 still has InternetA as its default route 
>> (even
>> though it has Node2 as another route), it tries to send all the pings via
>> the default route.
>>
>> What I (and Drew) have contend is that even this route should be removed
>> from the local route table preserving only the route via Node2. This I
>> proved by manually deleting the default route on Node1 via InternetA,
>> Pings
>> go through immediately!
>>
>> Thanks,
>> Rajesh.
>>
>> On 1/9/07, Rajesh Narayanan <(spam-protected)> wrote:
>>>
>>> Hi,
>>>
>>> Looks like we might be on to a bug here in the olsr dyn-gw. Yes, my
>>> opinion is the same as yours. The default route should be removed if the
>>> pings do not go through that route. Then I am sure it will behave
>>> correctly.
>>>
>>>
>>> For clarification to everyone regarding my setup.
>>> Node1 and 2 : Linksys WRT54G routers running the OpenWrt , WhiteRussian,
>>> Release RC4
>>> OLSR: compiled for RC4.
>>>
>>> For sake of simplicity please ignore the LaptopsA and B. My apologies
>>> its
>>> just confusing it. The laptops are being used mainly for accessing the
>>> Linksys routers console via SSH.
>>>
>>> Thanks,
>>> Rajesh.
>>>
>>>
>>> On 1/9/07, Drew <(spam-protected)> wrote:
>>> >
>>> > I've asked basically the same thing before, and never got a reply:
>>> > http://www.olsr.org/pipermail/olsr-users/2006-October/646474.html
>>> >
>>> > Good luck,
>>> > Drew
>>> >
>>> > Rajesh Narayanan wrote:
>>> > > Hi,
>>> > >
>>> > >
>>> > >                              *  *                         *  *
>>> > >                              |  |                          |  |
>>> > > InternetA--------->Node1 <-(OLSR)->  Node2---------->InternetB
>>> > >                              ^                              ^
>>> > >                              |                               |
>>> > >                              |                               |
>>> > >                Wired link laptop A        Wired link laptop B
>>> > >                   (HNA4)                        (HNA4)
>>> > >
>>> > >
>>> > > I got the dyn gw plugin up and running and configured as suggested.
>>> > > Well the problem is that :
>>> > > * Assume I remove the link the InternetA (cable disconnect to
>>> simulate
>>> > > link failure)
>>> > > * Node 2 gets notified that a route does not exist to the internet
>>> > anymore
>>> > > * But the Route to internetA still remains in Node 1!!!
>>> > > * Node1 tries to route the pings (to google.com <http://google.com>)
>>> > > via this interface since its the default route
>>> > > * As the default route does not get removed (due to link failure)
>>> the
>>> > > OLSR link is not chosen.
>>> > > * If I manually remove the default route, only then the Node1
>>> forwards
>>> > > the pings via Node2 and out.
>>> > >
>>> > > What am I doing wrong? Any other config issues I need to look at??
>>> > >
>>> > > Thanks,
>>> > > Rajesh
>>> >
>>> >
>>>
>> _______________________________________________
>> olsr-users mailing list
>> (spam-protected)
>> https://www.olsr.org/mailman/listinfo/olsr-users
>>
>
>
>
> _______________________________________________
> olsr-users mailing list
> (spam-protected)
> https://www.olsr.org/mailman/listinfo/olsr-users 





More information about the Olsr-users mailing list