CONFIG
RadiusNT 1.16.49b, running as a service on NT4.0/SP3, dedicated auth/acct
server
PROBLEM
Users couldn't authenticate, found RadiusNT service had stopped. It
wouldn't restart, returned a general system error ("1067: the process
terminated unexpectedly"). Command line would also not start the app,
running radius -x15 reported "ODBC: SQLConnect Error: [Microsoft][ODBC
Microsoft Access 97 Driver] General error, failed to get the expression
service."
CLUES
This system had been running without a glitch for several months, problem
seemed to occur out of the blue. But there *was* one recent change,
coincidentally on the afternoon just before the trouble started. We were
setting up an AT scheduled process to run a batch file to compact our
auth/acct db ... it was to stop RadiusNT, then compact the db, then restart
RadiusNT, simple enough. But it was pausing (and so failing) at the compact
step because of Microsoft's very annoying little "Mr. Paperclip," which we
couldn't find anywhere to disable from running at startup. One of our techs
thought this was part of the wizards option at installation, so he
reinstalled Access97.
TROUBLESHOOTING
Since the problem started the following morning, we associated these events
together and spent quite a lot of time investigating possible ODBC
problems, reinstalling drivers, removing and installing Access97 again,
removed and reinstalled the RadiusNT service, tried running it as a command
line app, double checking registry values, etc etc. Searched through all of
RadiusNT's online archives (very handy Dale) as well as a general Net
search for ODBC problems - tried a couple of suggestions we found, but all
with no change. Also, when attempting to repair the db via the ODBC DSN
manager, the same error message appeared as noted earlier ("failed to get
the expression service"). However, the db would successfully repair and
compact via Access itself.
SOLUTION
Dale sent an e-mail recommending installing version 2.5, since v1.x is no
longer supported and he knew that ODBC issues were handled a bit
differently now. My preference had been to try to fix the original problem
and get back to square one first, since I don't like to make several system
changes all at the same time - that's living dangerously! But we had run
out of airspeed, altitude, and ideas so decided what the hell, nothing to
lose.
It worked! Immediately after removing 1.16 and installing and configuring
2.5 (plus the update), the RadiusNT service would start again. We then
started receiving connection attempts that were failing, with a confusing
error message about "no attributes," even though RadAttributes were already
there in our previous db. Anyhow, there were a few new tables to create and
several new fields to add to preexisting tables - although there did not
seem to be any obvious relationship to the attributes error, that problem
nevertheless cleared right up as soon as we finished with the db updates -
normal connections resumed immediately (came flooding in is actually more
like it).
RECAP
The good news is that all is running great again! I still have a question
mark about what actually happened to cause the problem in the first place,
but that may never be known. Would a reinstall of 1.16 have solved it?
Maybe, but by the time we reached that point it didn't make sense not to
move up to 2.5 (if that had not worked we would have backtracked).
Thanks to Dale for the online support and suggestions, even on a weekend
(when else do things go wrong?), and for the new version. Hope these notes
might assist someone else...
Regards,
Ted Olson
OCS Software