[nos-bbs] trace file delays

Michael Fox - N6MEF n6mef at mefox.org
Thu Nov 4 10:10:55 EDT 2010


Maiko,

 

Here's some more info on the problem with delayed writing to trace files.
(You're probably tired of hearing from me on this, but I figured you can
always use more specific data).  J

 

I have a test machine set up with 3 ports but one has no radio connected.
It's sending beacons every 30 minutes but, of course, hears nothing on that
one port.  So the trace file consists of only outgoing beacon packets.
Trace level is set to 0111.  The file is 12 hours behind!

 

Another port sits on a frequency shared by a couple other BBSs.  At night,
there are beacons from all three machines, plus some low levels of traffic.
Trace level is also set to 0111.  This morning, when I check the machine,
the file was 4 hours behind!

 

I started a ping to one of the other BBSs on that port:  "ping xxx 56 10000"
for a 10,000 ms gap between pings.  After a while, I got the idea of using
tail -f.  So I opened another terminal session and did a "tail -f" on the
trace file to see when it was being written.  It was written after packets,
26, 34, 46, 54, 63, .  or about every 9-10 pings or about every 90-100
seconds.  

 

Now, the machine can capture tcpdump or tshark trace output on-the-fly, at
Ethernet speeds.  So it should be able to easily write trace output for a
1200 baud connection as it happens - just like it already does to the
screen.

 

I understand that some implementations of JNOS may be running on very old
machines, although I find it difficult to believe that they wouldn't be able
to keep up with 1200 baud.  Anyway, let's assume they can't.  Perhaps it
would be possible to add a write delay parameter to the config file?  Set it
to zero and it sends info to the file system at the same speed it writes to
the screen.  Suggest a value that would slow it down to its current crawl -
perhaps even make that the default, for backward compatibility.

 

I hope that's useful.

 

Michael

N6MEF

 

P.S.  A couple of us have been running the "I" release and no problems so
far.  Of course, we haven't had it running for 25 days yet, so can't comment
on the timer fix yet.

 

 

 

 

 

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.tapr.org/pipermail/nos-bbs_lists.tapr.org/attachments/20101104/5e54240c/attachment.html>


More information about the nos-bbs mailing list