Dave Dykstra [[EMAIL PROTECTED]] writes:
> No, I tried pulling this file from several different machines, and other
> files work. Also, a tcpdump analysis showed TCP getting confused and
> netstat shows stuff in the send queue on the sending machine while it
> hangs. I suspect this one machine has some special networking hardware or
> an experimental driver or some debugging code.
Are you sure TCP is confused, or just that it is continuously trying
to send the data? One possibility - remote, but believe it or not
it's happened to me in the past - is pattern sensitivity on the
physical circuit over which this machine is connected. If something
is misconfigured in the CSU/DSUs along the path (e.g., a B8ZS hop
obscuring a misconfigured AMI hop) you may run into ones-density
problems or other issues that can result in a loss or corruption of
data. In such a case an interesting test is to determine the actual
bytes that rsync is trying to send for those 60 but via some other raw
TCP mechanism and see if they still fail to get through. Or just
putting the circuit through some circuit standard stress tests can
sometimes shake out stuff like this.
Generally though, stuff like this is most likely to be ticked by runs
of common data patterns (in one case, it was the an IBM text logo in a
PC file - the old block character one - that was lots of the same hex
block character for each segment of the IBM logo :-))
-- David
/-----------------------------------------------------------------------\
\ David Bolen \ E-mail: [EMAIL PROTECTED] /
| FitLinxx, Inc. \ Phone: (203) 708-5192 |
/ 860 Canal Street, Stamford, CT 06902 \ Fax: (203) 316-5150 \
\-----------------------------------------------------------------------/
