[Voyage-linux] Re: [Madwifi-users] Weirdness - 12 second dropouts every 4 minutes

[Edwin Whitelaw]

Update on things...

Seems it's *not* a wireless or madwifi issue.  I have determined that 
the zebra daemon on the NRVU box is resetting but it does not correlate 
to any observed RF variations or link status, etc.  Making my life 
further confused, I am running quagga (zebra +ospf) on the other two 
WRAPs and they are working fine.  Everything is the same that I can 
determine so far.

Thanks for everyone's suggestions.

Edwin

Edwin Whitelaw wrote:
> Re my original post, it's not the eth0 port that seems to be the 
> problem.  The Hubbell relay unit does have an AP hung off the eth0 
> port but there is almost no traffic on it and besides, my tests 
> clearly show ath0 to be the interface that drops.  I can ping the eth0 
> port from the CRC box while pinging through the ath0 to ath1 on 
> Hubbell and the first ping will continue while the second one shows 
> the periodic interruptions.
>
> The eth ports on the two end units do see all the traffic but at this 
> point, I would not call it "heavy".  The eth port on the CRC end is 
> connected directly to a Cisco switch (vlan into a router elsewhere) 
> and the one on the NRVU end connects to a 100Base switch. The wireless 
> link shows 18-19 mbs under iperf tests and my real world load rarely 
> breaks 3 mbs end-to-end though I do hope to load it more soon with 
> more customers.  :-).
>
> In answer to your question, Mike...
> natsemi dp8381x driver, version 1.07+LK1.0.17, Sep 27, 2002
>  originally by Donald Becker <becker at scyld.com>
>  http://www.scyld.com/network/natsemi.html
>  2.4.x kernel port by Jeff Garzik, Tjeerd Mulder
>
> I am truly stymied on this one but but think the 239 sec interval and 
> the 12 sec duration are major clues.  What happens every 4 minutes 
> that would cause a radio to not pass traffic?  What is the 
> significance of the occasional 189 and 50 sec sub interval?
> On a possibly related note, I'm seeing a fair number of crcerrs when 
> running athstats on the NRVU-Hubbell link or should I say, what 
> constitutes a lot of crcerrs?.  Interestingly enough, there are 
> considerably more errors for outbound traffic than inbound though 
> iperf numbers are the same going both directions.  In any case, the 
> performance is wonderful *except* for the dropouts so unless the 
> crcerrs are related to that, I'm not terribly concerned.
>
> -----crcerrs from athstats1 on nrvu---------
> nrvu:~# athstats 1
>   input   output altrate   short    long xretry crcerr  crypt  phyerr 
> rssi rate
> 4444516  3604549    1199       0  158591   9148  48327      0       
> 3   30  54M
>      53       49       0       0       0      0      0      0       
> 0   31  54M
>      54       42       0       0       0      0      0      0       
> 0   30  54M
>     176      107       0       0       0      0      1      0       
> 0   30  54M
>      80       60       0       0       0      0      0      0       
> 0   28  54M
>      67       59       0       0       0      0      0      0       
> 0   31  54M
>      81       55       0       0       0      0      0      0       
> 0   30  54M
>     100       90       0       0       1      0      1      0       
> 0   30  54M
>     124       92       0       0       0      0      0      0       
> 0   30  54M
>     186      133       0       0       3      0      1      0       
> 0   30  54M
>     168      132       0       0       2      0      1      0       
> 0   30  54M
>     205      163       0       0      11      0      4      0       
> 0   30  54M
>     109      118       0       0       4      0      1      0       
> 0   30  54M
>     142      131       0       0       0      0      0      0       
> 0   30  54M
>     116       88       0       0       2      0      1      0       
> 0   30  54M
>     105       70       0       0       1      0      1      0       
> 0   30  54M
>      44       47       0       0       2      0      0      0       
> 0   30  54M
>      19       17       0       0       0      0      0      0       
> 0   30  54M
>      69       58       0       0       0      0      1      0       
> 0   31  54M
>      83       80       0       0       0      0      0      0       
> 0   31  54M
>      87       67       0       0       0      0      0      0       
> 0   30  54M
> --------end of athstats-----
>
> Edwin
>
> PS.  Testing again tonight and there seems to sometimes be longer 
> "good" periods than the 239 sec but the pauses are still exactly 12 
> seconds long and still unacceptable.  :-(  -e
>
>
> Michael Renzmann wrote:
>> Hi.
>>
>> On Sun, 2005-11-20 at 11:20 -0800, Tom Sharples wrote:
>>  
>>> Don't know if this is apllicable to you, but we have experienced 
>>> problems using the ethernet port on WRAP boxes for some time; I 
>>> haven't noticed the specific pattern you're describing but every 
>>> WRAP box we've ever deployed has had a tendency to hang on heavy 
>>> sustained ethernet traffic, usually once every few days.
>>>     
>>
>> Which driver do you use for the DP83815/6 chip that drives the ethernet
>> interfaces in the WRAPs?
>>
>> Bye, Mike
>>
>>
>>
>> -------------------------------------------------------
>> This SF.Net email is sponsored by the JBoss Inc.  Get Certified Today
>> Register for a JBoss Training Course.  Free Certification Exam
>> for All Training Attendees Through End of 2005. For more info visit:
>> http://ads.osdn.com/?ad_id=7628&alloc_id=16845&op=click
>> _______________________________________________
>> Madwifi-users mailing list
>> Madwifi-users at lists.sourceforge.net
>> https://lists.sourceforge.net/lists/listinfo/madwifi-users
>>   
>

-- 
<=+=+=+==+=+=+==+=+=+=+=+=+=+=+=>
Edwin Whitelaw, P.E.
New River Valley Unwired, LLC
2200 Lonesome Dove Dr
Christiansburg, VA 24073
540-239-0318