It seems many other Solaris 2 users have encountered this problem. Sun is aware

of the problem - apparently there are 2 patches for the problem - 101533 &

101615. The trouble is Sun haven't yet released these patches!

The solution (!) that several respondants mentioned, and the one I am currently

using is to run a little daemon called utmpd. It runs in the background and

checks the utmp / wtmp files every 15 seconds and fixes them if they are

inconsistent. It is available by anonymous ftp from "ftp.chpc.utexas.edu"

in a directory called "packages/sunos5" and is called "utmpd.tar".

 The only problem for my setup is that it also zaps the X session entries that

are put in by 'sessreg', presumably because their controlling process only

exists for a brief time.

 So for the time being running utmpd seems the best solution.

Thanks to :-

=========

grs@com.claircom (Gregg Siegfried)

Jason L Tibbitts III <tibbs@edu.uh.tcamc.sina>

adap@edu.andrews (Edsel Adap)

David Stewart <D.A.Stewart@uk.ac.ed.ee>

jose@es.ehu.lc.we (Jose M. Alcaide)

John Caruso <caruso@edu.umd.paradiso>

Linda Cygan <cygan@com.wpm>

Linda Pratt <lcp@uk.ac.liv.ph.ns>

----- Begin Original Message -----

Systems: SPARCserver 1000, dual processor, 192Mb, prestoserve, Solaris 2.3

         SPARCstation 1+, 24Mb, Solaris 2.3

Problem:

========

 On both systems (and on the local Computing Centre's SS1000 running Solaris

2.3), there is a problem with the login records. I presume it's got to do with

the /var/adm/utmp, wtmp files.

 The symptoms are that when a user logs out, their pseudo-terminal entry is not

cleared. The most common effect is that 'finger <user>' shows them still logged

in. Usually 'finger' (with no arguments), 'w' and 'who' do NOT show the user

being logged on. However occasionally, I see output from who (or 'w' & 'finger')

like the following:

root console May 2 19:33

arana pts/5 May 3 10:42 (grouse)

nmurray pts/1 May 2 19:56 (cs100-w)

nmurray asterix:0 May 2 19:59

nmurray pts/0 May 3 09:47

watson pts/0 May 3 10:01 (cs100-w)

Note the two entries for pts/0. This means that many programs misread the user

on pts/0 as me (nmurray). I believe this may also have something to do with

NIS+ error messages I've been getting (which I posted a message about several

weeks ago, and have still mot resolved).

Patches installed: 101331-03 101316-01 101344-05 101347-01 101384-05 101318-41

                   101500-02 101317-08 101329-13 101327-05 101411-04 101448-01

                   101416-02 101494-01 101513-02 101543-01 101514-01 101512-07

                   101415-02

Does anybody have an ideas about this?

Thanks in advance,

Nick Murray

Computer Officer

Department of Computing Science

Aberdeen University

----- End Original Message -----