SUMMARY: strange NIS+ problem

From: Jenny Partusch (jenny@tan.unl.edu)
Date: Fri Mar 12 1999 - 11:01:12 CST


Months and months ago, I asked for help with a problem we were having with
our 2.6 NIS+ clients. We suspected that we had some corrupt tables or
something of that nature because of the strange problem that was
occurring. Nobody had an answer that led us to a solution at that time,
but a friend recently stumbled across the same problem (and a solution!)
on SunSolve.

What was happening? Every few days, some of our 2.6 machines would start
having problems with authenticating NIS+ users. We could still read some
of the NIS+ tables, but "niscat passwd.org_dir" would give an error to the
effect that it couldn't talk to the NIS+ master. I tried all
kinds of things like removing the credentials for problem hosts,
sys-unconfig and reconfiguring, and readding the machine as a client, all
with no effect.

After much troubleshooting, I learned that stopping and restarting keyserv
would fix the problem, but three days later at 11:00 pm, the problem would
recur. We were puzzled because the only cron jobs that took place on any
of our machines at 11 were ones that cleaned out /tmp. Then we realized
that they got rid of any files over 2 days old and thus removed the
.rpc_door directory every three days after it was created. We didn't have
the cron on all of our 2.6 machines and the 2.4/2.5.1 machines worked
fine.

Since modifying the cron to leave files owned by root, we haven't had any
further problems with NIS+. Thought I'd send a belated summary in case
somebody else runs into this problem.

Jenny

--
Jenny Partusch                |   CALMIT
jenny@tan.unl.edu             |   113 Nebraska Hall
UNIX Systems Administrator    |   University of Nebraska - Lincoln



This archive was generated by hypermail 2.1.2 : Fri Sep 28 2001 - 23:13:16 CDT