Path: utzoo!attcan!utgpu!watmath!att!rutgers!ucsd!tut.cis.ohio-state.edu!purdue!decwrl!hplabs!hp-sdd!ucsdhub!sdcsvax!odin.ucsd.edu!mikulska
From: mikulska@odin.ucsd.edu (Margaret Mikulska)
Newsgroups: comp.unix.wizards
Subject: Re: Why does my /etc/syslog rack up CPU time?
Message-ID: <7464@sdcsvax.UCSD.Edu>
Date: 15 Nov 89 23:22:56 GMT
References: <3531@chorus.fr>
Sender: nobody@sdcsvax.UCSD.Edu
Reply-To: mikulska@odin.ucsd.edu (Margaret Mikulska)
Organization: University of California, San Diego
Lines: 37

In article <3531@chorus.fr> rd@chorus.fr (Roland Dirlewanger) writes:
>
>In article <110@cupcake.sal.wisc.edu>, jwp@larry.sal.wisc.edu (Jeffrey W Percival) writes:
>
>> We have a MicroVax running Ultrix, and the /etc/syslog process seems to
>> be locked in some kind of deadly embrace...  it is building up CPU time
>> as if it is CPU bound.  ps(1) shows accumulated times like 675.23, and
>> shows the time to be increasing at nearly real time.
>
>I've already seen this on SunOs a 4.0 system.
>
>The very same problem appears if the syslog server for a host is itself.
...
>Check your /etc/syslog.conf for lines such as  :
>
>	something.level			@loghost
>	foo.bar				@localhost

Under SunOS 4.0, this is a bug: LOGHOST is not correctly defined by 
syslogd on loghost machines. The workaround consists indeed of 
correcting /etc/syslog.conf on loghost machines: you should add the 
following line:

define(LOGHOST, 1)

as the _first_ line of this file. It works, I know from my own 
experience on some Sun-4's that really did almost nothing but ran
syslogd.


Margaret Mikulska

Univ. of Calif., San Diego
Inst. for Nonlinear Science

mem@inls1.ucsd.edu
ucsd!inls1!mem