[Battlemesh] Internet at the venue works :-)
dave.taht at gmail.com
Wed Aug 5 10:45:37 CEST 2015
On Wed, Aug 5, 2015 at 10:33 AM, Mitar <mitar at tnode.com> wrote:
> It was combination of some hardware issues and human issues:
> - one fiber connection we have failed on one remote roof
> - one ethernet jack was not done well and was losing packets
> - one remote router had an ethernet port failure
> - because of the ethernet port issue on OpenWrt & Ubiquiti (reported a
> week ago) we have some things sub-optimally done through bridging, so
> less fail-over and best route selection (so we had to manually drop some
> links, because OLSR does not support route filtering like Babel does)
> - so because of this initial hardware issues some people were trying to
> be helpful and were modifying cables on the table and other stuff, which
> made things worse and hard for us to debug
> - and later on when we fixed all hardware issues, there was still your
> QoS configuration from a day ago
> So, yea, mostly bad luck. :-) And no WiFi issues, just everything else.
> :-) WiFi links worked well from the beginning. :-)
> So currently the venue has a nice uplink. Internally, the spectrum is
> crowded, but we cannot really do much about this if we do not want to
> use too much spectrum away from testing. I am not sure what channels
> does the testbed need. We put APs on 1 and 13 for now.
1) Presently the link I put in in the conference room (babel-lecture)
is on backhauled channel 36 and 11 for the ap, HT20. It is natted off
from the main net. If you want to assign some 10 space or suggest
something different back there?
2) I would imagine that most laptops here have 5ghz capabillity,
perhaps we could dedicate another 5ghz channel for conference use, and
agree to keep the rest clear for testers? 40? 44?
I suggested using 161 and 165 for this:
- and 1 and 11. perhaps 6 and 11?
What are the other tests going to use?
>> I would certainly love to have a blow by blow account of what has gone
>> wrong so far and how it was fixed so far.
>> (I know I am at fault for semi-fixing it with qos one day, and having
>> it be too slow the next)
>> I note that at least some of the problems we have had were by having
>> too much load on the system... I have observed latencies well in
>> excess of 2 seconds over the course of the last few days, in addition
>> to major loss.
>> and if we all come, it might break again. (Look! a pool! Wheee!)
>> Are there any statistics being collected on the path from here to there?
>> If you like I could put up a mrtg snmp sampling various bits of the
>> path, and also smokeping is of use.
>> On Wed, Aug 5, 2015 at 9:11 AM, Mitar <mitar at tnode.com> wrote:
>>> Report from the venue: it works 30/15 Mbit/s. :-)
>>> You can all come. :-)
>>> Battlemesh mailing list
>>> Battlemesh at ml.ninux.org
> Battlemesh mailing list
> Battlemesh at ml.ninux.org
worldwide bufferbloat report:
What will it take to vastly improve wifi for everyone?
More information about the Battlemesh