Insanely high "Cached" value

Insanely high "Cached" value

[Steven Walter]

My system has been running a little over twelve days now, and I just
noticed that the "Cached" value in both 'free' and /proc/meminfo is
insanely high.  This wasn't the case the last time I checked, which was
probably a day ago.

Just before checking it this time, I ran a "du -s *" in /usr, which
generated a lot of I/O, as it to be expected.  Perhaps the large amount
of I/O has uncovered a bug of some sort?

This is kernel 2.4.13 (hopefully it's not something that's already been
reported and fixed; I haven't seen it if is has) patched with ext3, kdb,
lm_sensors, and the pre-empt patch.  Seems likely to be only a simple VM
problem, however, and an asthetic one at that.
-- 
-Steven
In a time of universal deceit, telling the truth is a revolutionary act.
			-- George Orwell
Freedom is slavery. Ignorance is strength. War is peace.
			-- George Orwell
Those that would give up a necessary freedom for temporary safety
deserver neither freedom nor safety.
			-- Ben Franklin
He's alive.  He's alive!  Oh, that fellow at RadioShack said I was mad!
Well, who's mad now?
			-- Montgomery C. Burns

Re: Insanely high "Cached" value

[Andrew Morton]

Steven Walter wrote:
> 
> My system has been running a little over twelve days now, and I just
> noticed that the "Cached" value in both 'free' and /proc/meminfo is
> insanely high.  This wasn't the case the last time I checked, which was
> probably a day ago.
> 
> Just before checking it this time, I ran a "du -s *" in /usr, which
> generated a lot of I/O, as it to be expected.  Perhaps the large amount
> of I/O has uncovered a bug of some sort?
> 
> This is kernel 2.4.13 (hopefully it's not something that's already been
> reported and fixed; I haven't seen it if is has) patched with ext3, kdb,
> lm_sensors, and the pre-empt patch.  Seems likely to be only a simple VM
> problem, however, and an asthetic one at that.

It's an ext3 bug.  Harmless, fixed in the (ext3-enriched) 2.4.15-pre2.

-

Re: Insanely high "Cached" value

[Martin Knoblauch]

> Re: Insanely high "Cached" value
> 
> From: Andrew Morton (akpm@zip.com.au)
> Date: Sat Nov 10 2001 - 00:17:01 EST
> 
> 
> Steven Walter wrote:
> >
> > My system has been running a little over twelve days now, and I just
> > noticed that the "Cached" value in both 'free' and /proc/meminfo is
> > insanely high. This wasn't the case the last time I checked, which was
> > probably a day ago.
> >
> > Just before checking it this time, I ran a "du -s *" in /usr, which
> > generated a lot of I/O, as it to be expected. Perhaps the large amount
> > of I/O has uncovered a bug of some sort?
> >
> > This is kernel 2.4.13 (hopefully it's not something that's already been
> > reported and fixed; I haven't seen it if is has) patched with ext3, kdb,
> > lm_sensors, and the pre-empt patch. Seems likely to be only a simple VM
> > problem, however, and an asthetic one at that.
> 
> It's an ext3 bug. Harmless, fixed in the (ext3-enriched) 2.4.15-pre2.
> 

 Hmm. Are you sure it is ext3 only? I see the same (coming and going, no
real harm) on 2.4.13-ac4+preempt without having EXT3 enabled. Also
happens with 2.4.13 plain.

# CONFIG_EXT3_FS is not set
CONFIG_EXT2_FS=y


Martin
-- 
------------------------------------------------------------------
Martin Knoblauch         |    email:  Martin.Knoblauch@TeraPort.de
TeraPort GmbH            |    Phone:  +49-89-510857-309
C+ITS                    |    Fax:    +49-89-510857-111
http://www.teraport.de   |    Mobile: +49-170-4904759

Re: Insanely high "Cached" value

[Robert Love]

On Fri, 2001-11-16 at 04:54, Martin Knoblauch wrote:
>  Hmm. Are you sure it is ext3 only? I see the same (coming and going, no
> real harm) on 2.4.13-ac4+preempt without having EXT3 enabled. Also
> happens with 2.4.13 plain.

It is not preempt-kernel's fault.  There are two separate bugs, one in
ext3 and one in the cache reporting in -ac series.  Newer kernels fix
both problems.

2.4.13-ac7 and 2.4.15-pre5 are not affected ... 2.4.15-pre5 is probably
ideal, it merged ext3, too.

	Robert Love