To make a long story short:
MANY PEOPLE JUST DON'T REALIZE THE EFFECT THAT AN OUT-OF-DATE
CELLSERVDB CAN HAVE ON SITES SUCH AS OURS WITH LIMITED INTERNET
BANDWITDH. IT CAN **HURT**.
Some of the rest of the story:
At our site, we upgraded our AFS servers from NeXTs to Alphas and Sparcs
last summer. As a part of this upgrade, we changed the names and addresses
of all three of our database servers. This was on August 18, 1994.
However, we noticed a month ago that we had a rather large amount of
Internet-bound ICMP traffic being generated by our old NeXT database
servers. We thought it was something related to DNS or NeXT's NetInfo
(which also run on those three machines). However, upon closer inspection,
we discovered that it was "Destination Port Unreachable" messages regarding
UDP port 7003 - the AFS Volume Location server.
With a little more investigation of our internet connection, we discovered
that this traffic consisted of anywhere from 150-500 "bad" requests per
minute; I wrote a Perl wrapper to Solaris's "snoop" program that counted
UDP port 7003 traffic and ICMP traffic for these three machines. One run
today provided the following:
Requests: 419 Bytes: 63075 Bytes/sec: 1051.25
Bits/sec: 8410.00 Percent of 56KB: 14.67
Note the "percent" field. Since we have only a 56KB link to the Internet,
this bad traffic is taking up an **UNACCEPTABLE** amount of our bandwidth.
I understand that many (most?) sites have automagic methods for getting
that latest and greatest CellServDB from Transarc. However, I also
understand that it is the decision of each machine's administrator whether
or not to use these methods.
I have talked with Transarc about this, and they are currently working to
verify the correctness of their CellServDB.export file. They plan to have
site contacts verify their cell's entry in the .export file and send any
corrections. After two weeks, they will post a plea to all admins to
update their CellServDB file. Transarc also plans to supply scripts which
will help to:
- verify the CellServDB file.
- add and remove mount points from /afs to match the .export file.
Look for information from them on this issue in the near future.
To check your machine, just "grep 137.112 /usr/vice/etc/CellServDB". You
should see achilles, ajax, and hector. If you see zeus, hera, and red,
your machine is part of the problem. :-)
In the meantime, if you would like more information on how exactly we did
this study, free to drop me an e-mail. Or, look me up at Decorum 95
if you're going! :-)
- Mike
--
Mike Allard, Workstation Manager, Rose-Hulman Institute of Technology--KA9VDC
<[EMAIL PROTECTED]> - #include <std/disclaimer.h> - NeXTmail Ho!
Murphy's Law is not recursive; washing your car to make it rain does not work.
"The steady state of disks is full." -Ken Thompson
