eric

Yesterday I installed Power Pack 2006 on a PIII system
Ever since then i notice there is a disk access every few
seconds, usually about every 2 to 5 seconds or so.
Whats doing that?

Also, what is up with xorg being an unsupported pre-release version?
Thanks
Eric

robert m. riches jr.

Top theories are Kat or ext3 (or other) journal flushing.
IIUC, general opinion is Kat should simply be uninstalled.
Journal flushing, if it is the cause, can probably be tuned
if it's a serious problem. Usually, default settings for
journal flushing are tolerable in most situations.


There are a few errata items about xorg. (They are the
major reason I haven't yet worked up the courage to install
2006 myself.) The good news is there's an update in the
works that appears to go completely over to 6.9.

--
Robert Riches
spamtrap42@verizon.net
(Yes, that is one of my email addresses.)

eric

I should have noted that I'm in console mode (init 3) when
I observed the disk activity (it also is there with kde running). I dont use
the GUI much on this machine and most of the time it sits in run level 3. I
un-installed kat anyway but i still see the cyclic disk accesses while no
one is even logged on and its sitting at the run level 3 login prompt. This
is going to kill the drives eventually and it certainly wont let them spin
down and sleep.
How can i track down the culprit?
Thanks
Eric

robert m. riches jr.

I'd suggest you log in to the text console, _don't_ start X,
and run 'top' for a minute or two. If any processes show up
as running, hang 'em high. :-)

After that, do 'ps augx | less' to look at the processes on
the system, especially the ones started _after_ booting. If
there are any background KDE processes hanging around, you
could consider killing them to see whether they were doing
it.

What filesystems aer you using? Are you using ext3 or
another journaling filesystem? If so, it's fairly likely
that journal flushing is the cause of the activity.

What kind of situation do you have that causes you to
believe that periodic activity will cause unreasonable
shortening of the disks' lifespans? I have always kept my
disks spinning full-time without much difficuly. (Knock on
plastic. :-)

--
Robert Riches
spamtrap42@verizon.net
(Yes, that is one of my email addresses.)

sig sauer

Since you are running in init 3, it could be a remote machine
communicating with your linux system. Samba....NFS....CUPS....etc.
II notice this on my machine as well, init 5 or init 3.
As far as long term damage to the drives........I would be more concerned
about the HDD's having to spool up to speed everytime a request was made
to that machine. Typically, the motor that spins the platters has to
overcome all of the inertial forces everytime it gets a request, has to
spool up from sleep state, and then spool back down. It makes sense to me
that this would have more of a long term effect on disc longevity.
Conversely, it makes sense that if the hard drive is spooled up to speed
all of the time, that the disc motor is only keeping it up to speed,
(flywheel effect) and doesn't have to work as hard to do that as it does
to spool up the disc from a sleep state. Just a few thoughts. I have seen
servers (Novell file servers) up and running for an excess of 12 years,
never turned off, never sent into sleep mode.
Another consideration is heat. When a system is left in a full ON state,
heat remains fairly consistent within the range of ambient temperature
ranges that the computer is in. Most damage to the hard discs themselves
can occur from being heated and cooled repeatedly. This repetition of
heating/cooling causes expanding and contracting of the metal discs which
can eventually result in small microscopic cracks forming on the surface
of the hard discs. This results in data corruption and is probably why
many hard discs turn into paper weights.

The laws of physics will not be denied.

Sig

eric

I think it is the ext3 doing a 5 second flush.
see
#define JBD_DEFAULT_MAX_COMMIT_AGE 5
in file ./include/linux/jbd.h
This brings up the question: why is it blindly writing to disk every 5
seconds? shouldnt it be flushing only if there is data to be flushed?
Similar to a cache, only write back "dirty" cache data.

The thing about disk activity and lifetimes: I agree frequent spin up/spin
down cycles would be harmful. but if the system is only active say, once
every few hours for short periods, or maybe once in 12 hours, it seems you
would be extending the disk life if it could spin down and go to sleep.
Eric

nospamgc

<SNIP interesting answers>

I can't really help you with the OP 's question but this thread makes me
wonder how much extra disk access realtime virusscanning would have
shortened the lifespan of my disks /if I were running Windows an
(therefore) needed AV/. (Wich I am not).
In extent: the total cost of ownership consequences of needing realtime AV,
the extra energy needed because of power consumption as well as the lost
cpu cycles from realtime AV scanning adding to the total cost of ownership.
Heck, you may call this OT but in fact on a global scale were talking about
massive overhead here, are we not?
(Assuming realtime virusscanning adds to disk activity).

GC

eric

yeh, good point. I have to run windows at work and as long as i'm doing
something simple its ok. But the second i need to do a little more than
that its a gigantic roadblock. For example, today i ran something on xp and
one thing i wanted to know was: how long has it run so far? (as in 'top')
Well, lets see... ok task manager - nope that doesnt show me etc etc i was
stuck! stymied by such a simple task. You know whats really funny? I used
to be a diehard windows user - hated anything and everything unix. Now i
wont run anything but linux at home. linux has taught me more about windows
than i ever new before. Now thats weird isnt it?! but its true. Oh crap,
I'm rambling... sorry
Eric