[Karlnet] Strange retransmit problem

[Chris Conn <cconn@abacom.com>]

Hello to all,


I have been having a strange problem for quite a while.  I have spoken 
to Karlnet about this and they told me it was a rare occurrence and 
_maybe_ one other customer was seeing this but all data was inconclusive.

The problem:

A situation can arise intermittently (once a week, once a month, three 
times a day, no real pattern) where in a multi-point antenna, the base 
station will enter a mode where it is retransmitting thousands and 
thousands of packets to a satellite.  The transmit counter does not 
increase, only the re-transmit counter, and by the thousands at a time. 
 Upon looking at the Turbocell station entries, we notice that the 
transmit and retransmit counters seem to have reset back to zero for the 
satellite to which it is retransmitting, and you can watch the 
re-transmit counters increase at an alarming rate.  All other satellites 
are still visible, however no TCP/IP data can pass to any satellite, 
including the one to which the retransmits are occurring.  Therefore, 
the whole system is no longer useable.  The amount of throughput before 
or after can be high or low, it does not matter.  This will happen day 
or night, with or without customer usage.  After a few minutes (2 to 5), 
the whole thing comes back online and data starts to pass again.  It 
would seem that the base station jams on one satellite which no longer 
can acknowledge a transmission for some reason.

We have tried changing all hardware, using flashroms, AP1000s, new 
antennas.  In no case are we using amplification of any kind.  We also 
tried replacing the radio card in the base to a PRISM 2.5 200mW, and 
were initially happy however the problem eventually happened therefore 
it is not a radio firmware issue (since it is not Orinocco specific), 
and looks to be a software issue.

The Workaround:

We have found that setting the base from 11Mbps to 5.5Mbps will fix the 
problem in 3 out of the 4 setups that are doing this.  However, in one 
case we had to set a customer satellite all the way down to 1Mb in order 
to cease totally any form of this problem.  But, we obviously take a 
performance drop in the process.

Other info:

All bases are running 4.31, and the clients are running a mix of 3.88 
and 4.02.  I assume however that it has to do with the base since it is 
my understanding that polling should occur many times per second to each 
satellite and during these periods, several minutes go by before any 
satellite is polled other than the one the base seems to get stuck on.

Any ideas?

Thanks,

Chris