On 22.02.2014, at 14:20, Marek Lindner wrote:
On Thursday 20 February 2014 11:20:01 whangarei & opua wrote:
The setup is like this: whangarei is my router, it has one Wlan link to each of this 3 neighbor router, MAC in the grep, hostname as comment. This 3 routers are in one large building (across a park) but sometimes don't see each other. 2 of them have VPN connects. I've looked only to the next hop, by grep the lines beginning with the MAC of this neighbor routers. The intension was to get a better feeling of the link quality then with the 'iw dev wlan0-1 station dump' output. So i have observed this behavior...
Have attached the router output as a file, it should be more readable.
Unfortunately, the information you provided isn't comprehensive enough to comment on your case. Please take a look at bat-hosts to have meaningful names
At the time I capture the output I was not aware about bat-hosts. :(
in your originator table. Furthermore, we need a clear description of your setup for what the routing concerns. You can take a look at our routing scenarios page to get a feeling what we usually work with: http://www.open-mesh.org/projects/open-mesh/wiki/Routing_scenarios
Sorry, it's a little bit more complex setup, so I draw a diagram. Now my router mounted outside and the signal is much better, so only shorter times (<20 sec) without receiving update packets are displayed. Connection is good at the moment, but expect problems again after april, because of leaves on the trees, maybe together with rain.
A diagram is no must but certainly helpful.
As soon as we have a clearer understanding of how your setup looks like, you can explain what behavior you see and what you would expect, plus originator tables from each node involved.
Sorry, have no access to the neighbor router, it's a live environment.
Cheers, Marek
Don't think there is any routing-protocol issue, what you maybe expect. It's out of question, the WLan connection was very poor at this time. Links #1 - #3 (on the drawing) are up for approx. 90% over a week, while #4 maybe 50% and #5 + #6 maybe 30%. This means, my neighbors are only seeing each other sometimes directly, so I also expect there high packet loss, sometimes.
I have observed that a TQ entry from a direct neighbor will be present, also it has a last seen time of up to 180 sec, maybe more. My suggestion is to reduce the TQ in such a case over the time, because this link can't be reliable, at least not bidirectional. In my output you can see, it will not be happen now and there are very high last seen times.
Off cause, I have not monitored where traffic to outer (not directly connected) destinations go through... Because it's layer 2 I don't expect to see something like a default gateway, which is easy to monitor. As written earlier, I was not looking for routing issues.
It will be nice to be able to configure the dead timeout on the local router, maybe like it works with the hop_penalty. Also I miss some possibilities to do some kind of traffic management, for example decrease / increase the TQ of incoming updates on a neighbor basis on the local router. In my case I know, that one link is more reliable over the time than other ones, but have not found any way to priories it. (I don't talk about static routing.)
Have a nice day, Joe