Dear Tru64 people,
Thank you to Serguei Patchkovskii and Terry Horsnell who both
suggested a busy network might cause this problem. Terry said that NCD
X terminals can be very sensitive to network load (I have seen this on
older NCD terminals but the newer ones are much better).
The problem is now fixed, so I am a lot more relaxed, but still
puzzled. I already knew that a machine somewhere on campus was
flooding our DNS server with lookup requests (over a 100 per second),
but it was a while before I was able to track down the culprit and
unplug the network cable. It turns out that machine was running a
caching DNS server which forwarded requests it could not answer to
us. These requests were all for 238.42.231.193.in-addr.arpa.
As far as I can tell it wasn't network congestion which caused the
problem because the X terminals were all on fairly quiet
networks. And the problem also occurred when using command line tftp
from the same machine as the tftp server. Also the tftp transfers did
occasionally get off the ground and once started they seemed to finish
OK.
So my theory is that tftpd is doing a DNS lookup but not allowing
enough time for it to complete. If my theory is right the timeout must
only be a few seconds, because other lookups seemed to work reasonably quickly.
It is worrying that tftpd is so sensitive when all the other services
seemed to cope with the surge quite happily...generally speaking you
can throw a huge amount of network traffic at an Alpha before it
starts to show the strain.
Bob Vickers
--
==============================================================
Bob Vickers R.Vickers_at_dcs.rhbnc.ac.uk
Dept of Computer Science, Royal Holloway, University of London
WWW: http://www.cs.rhbnc.ac.uk/home/bobv
Phone: +44 1784 443691
Received on Mon Dec 13 1999 - 17:16:00 NZDT