If this isn't exactly what you wanted, please try our Search (there's a LOT of techy and non-techy stuff here about Linux, Unix, Mac OS X and just computers in general!):
From belal@sco.com Fri Apr 9 23:26:34 2004
Newsgroups: comp.unix.sco.misc
Path: nntp.TheWorld.com!news.mv.net!newsfeed.mathworks.com!news.kjsl.com!newsfeed.stanford.edu!enigma.xenitec.on.ca!not-for-mail
From: Bela Lubkin <belal@sco.com>
Subject: Re: High sar %sys
Resent-From: mmdf@xenitec.on.ca
by mail.ut.sco.com with SMTP; 9 Apr 2004 22:23:41 -0000
Submit-To: scomsc@xenitec.on.ca
Content-Type: text/plain; charset=us-ascii
Organization: [resent by] The SCOMSC gateway and Propagation Society
Content-Disposition: inline
Date: Fri, 9 Apr 2004 22:27:43 GMT
Message-ID: <20040409222743.GW24746@sco.com>
User-Agent: Mutt/1.4.1i-nntp3
To: scomsc@xenitec.ca
Mime-Version: 1.0
In-Reply-To: <20040409182305.GA11973@tkrh.demon.co.uk>
X-Nntp-Posting-Host: enigma.xenitec.on.ca
Originator: news@enigma.xenitec.on.ca (News subsystem owner)
References: <20040407203328.GA23978@tkrh.demon.co.uk> <20040408071014.GU24746@sco.com> <20040409182305.GA11973@tkrh.demon.co.uk>
Sender: news@enigma.xenitec.ca (News subsystem owner)
Precedence: list
Lines: 96
Xref: nntp.TheWorld.com comp.unix.sco.misc:168801
Tom Melvin wrote:
> Bela Lubkin made comment on Thu Apr 8 08:10:14 2004 :
> > Tom Melvin wrote:
> >
> > > I have a machine that is showing a high %sys in sar.
> > >
> > > Used cpuhog/iohog/memhog along with u386mon + top and nothing shows
> > > the offending process(es) - ps listings are normal. No sign of swapping.
> > > Resorted to brute force and killed process after process and sar
> > > still showing high %sys.
> > >
> > > Anyone any ideas?
> > >
> > > Running OSR 5.0.7 + Update Pack 2
> > > Single Pentium P4 + 512Mb Ram + Adaptec Raid
> > > HyperThreading Enabled.
> >
> > Link scodb into the kernel ('Y' in /etc/conf/sdevice.d/scodb; relink;
> > reboot). While the problem is happening, do some ad hoc kernel
> > profiling by repeatedly breaking into scodb. At a text console
> > multiscreen, hit Ctrl-Alt-D. At the resulting scodb prompt, "b" (a
> > short alias for "stack"), then "q". Repeat many times. Get a feel for
> > what the kernel is doing when you expect it to be idle.
>
> Is there another way to 'access' scodb - I am remote so will either be
> using a telnet session or dial-up modem - while the modem is a character
> based tty line CTRL X or ALT-CTRL-D does not want to operate.
There is a way; whether it can be used in this case, you will have to
determine.
The way is: have someone local to the machine hook its COM1 serial port
up to a serial port on a neighboring machine. Let's call the machines
"victim" and "buddy". If buddy is running Unix, what you're going to do
is connect to it -- doesn't matter if this is via ssh, telnet, dialup,
whatever -- and `cu` out the port that's connected to victim. If buddy
is running something else then you'll have to figure this part out.
At some point you have to reconfigure victim to use a serial console.
If you are able to get a serial login (coming in through buddy), you can
do this all by yourself, remotely. Otherwise you'll need at least a
little on-the-spot assistance.
To change victim to a serial console for the next and subsequent
reboots, edit /etc/default/boot, add a line that reads:
SYSTTY=1
To change it for a single reboot, use a local assistant. Reboot the
machine. Have them type at the boot prompt,
Boot
: systty=1
You should then see a boot prompt on your serial connection.
If scodb is already linked into the kernel, you can change the console
"on the fly", with the help of a local assistant. Make sure they're
looking at a text multiscreen (have them hit Ctrl-Alt-F1). Then have
them hit Ctrl-Alt-D and enter:
scodb> dbtty(1)
A scodb prompt should appear on your serial connection. 'q' to get the
system back in gear. Login, monitor for the %sys-too-high condition;
once you see it, start doing ad hoc profiling.
In all cases, configure the comm program that will be talking to the
serial console for 9600-8-N-1.
Actually there may be an even easier way. Rereading what you wrote,
apparently you're dialing in directly to a modem on victim. In order
for that to work as a serial console, the modem needs to be on COM1
(tty1A). They've probably got it on some sort of intelligent multiport
board. Move it onto COM1 and try again.
Well, ^X still won't work, because at that point the console is the
video board, not COM1. So you still need to either (1) reboot with
"SYSTTY=1" in /etc/default/boot, (2) reboot and enter "systty=1" at the
boot prompt; or (3) break into scodb on the video console, enter
"dbtty(1)" to move the console to COM1.
I prefer the buddy system. In particular, I like to set up two OSR5
systems side by side with COM1 hooked to COM1. Then either of them can
have its console temporarily changed to COM1 (using "SYSTTY=1" in
/etc/default/boot). ssh into buddy and `cu -ltty1a direct` to get
access to victim's console. Even better: install GNU screen on buddy.
Run it. Within screen, `cu -ltty1a direct` to victim's console. Now
you have a persistent connection to the console, won't be dropped if
your serial/telnet/ssh connection in to buddy goes down. It'll only be
dropped if buddy itself goes down.
>Bela<
Enter your email address for automatic notification of new posts here
(be sure to whitelist 'feedburner.com' if you use spam filtering)
| Views for this page | ||||
|---|---|---|---|---|
| Today | This Week | This Month | This Year | Overall |
| 1 | 4 | 5 | 5 | 360 |
/Bofcusm/2453.html copyright 1997-2004 (various authors) All Rights Reserved
Have you tried Searching this site?
Unix/Linux/Mac OS X support by phone, email or on-site: Support Rates
This is a Unix/Linux resource website. It contains technical articles about Unix, Linux and general computing related subjects, opinion, news, help files, how-to's, tutorials and more. We appreciate comments and article submissions.

Click here to add your comments