hard disk being accessed every 4 sec.

[Paul Coleman]

After updated fresh install the hard disk is being accessed every 4 seconds
constantly.

0000:00:00.0 Host bridge: VIA Technologies, Inc. VT8601 [Apollo ProMedia] (rev 05)
0000:00:01.0 PCI bridge: VIA Technologies, Inc. VT8601 [Apollo ProMedia AGP]
0000:00:11.0 ISA bridge: VIA Technologies, Inc. VT8231 [PCI-to-ISA Bridge] (rev 10)
0000:00:11.1 IDE interface: VIA Technologies, Inc.
VT82C586A/B/VT82C686/A/B/VT823x/A/C PIPC Bus Master IDE (rev 06)
0000:00:11.2 USB Controller: VIA Technologies, Inc. VT82xxxxx UHCI USB 1.1
Controller (rev 24)
0000:00:11.3 USB Controller: VIA Technologies, Inc. VT82xxxxx UHCI USB 1.1
Controller (rev 24)
0000:00:11.4 Bridge: VIA Technologies, Inc. VT8235 ACPI (rev 10)
0000:00:11.5 Multimedia audio controller: VIA Technologies, Inc. VT82C686 AC97
Audio Controller (rev 40)
0000:00:12.0 Ethernet controller: VIA Technologies, Inc. VT6102 [Rhine-II] (rev 51)
0000:01:00.0 VGA compatible controller: Trident Microsystems CyberBlade/i1 (rev 6a)

It is unclear which process is responsible for this constant access. Top
indicates that multiload-applet is the number two time hog.

[Jeff Fortin Tam (kiddo)]

I have exactly the same problem on both a Hoary desktop (which was dist-upgraded
from warty) and on a Hoary laptop. I would really, really, really like this to
be fixed (or at least know why it is doing this) because it prevents hard drive
sleeping ~~ on the main drive only, secondary mounted drives are not affected ~~
so it is noisy (old drive), and not good for the drive either I think.

When entering battery mode on the laptop, the problem stops. When plugging it
back, it comes back.

[Steve Reid (sreid)]

I've managed to solve this on my box (Hoary) and wanted to share...

gnome-cups-icon seems to be the culprit, or at least the worst offender. I've
killed it off (System -> Preferences -> Sessions, Current Session tab, select
gnome-cups-icon, Remove) and my computer is a lot quieter now. I don't have any
printers set up so it's useless to me anyway.

It's possible other processes could cause the same problem, so YMMV.

I was able to track this down with "kill -stop", listening, then "kill -cont".
Half at a time (binary search) instead of one by one.

[Lakin Wecker (lakin)]

Can any of you retest this bug on a newer version of Ubuntu, preferably Dapper, but breezy will do. If the bug persists in these versions please re-open this bug and provide us with the version of the software which is causing this bug.

[Laurent Bigonville (bigon)]

I have the same issue in dapper

[Freakcode (freakcode-deactivatedaccount)]

Same issue in Dapper for me, too.

There is HDD activity only when dbus service is started, so he is the guilty for Dapper.

This is a SERIOUS bug, because I cant kill dbus without breaking GNOME functionality. There is low OpenGL performance too, when I got HDD I/O, the OpenGL "frezees" for a moment. I got I/O in 3-4 seconds, so OpenGL is freezing "way too much".

Im using kernel 2.6.15-23-k7 and nVidia installer drivers version 8762, on a Athlon 64 3000+ and Asus K8N motherboard.
The problem occurs with 386 kernel and original ubuntu nvidia drivers too.

[padre999 (patrickheeren)]

Hi everybody.

I have the same annoying problem on my Desktop since a while. This is on Ubuntu Dapper 6.06.
But I have to say that the write access to the disk is every 5 secs and not 4 secs on my computer. They are blocks of about 30-70 kb written to the disk as GkrellM tells me. There are also some threads in the Ubuntu forums dealing with the issue:

http://ubuntuforums.org/showthread.php?t=36976
http://www.ubuntuforums.org/showthread.php?p=660011
http://forum.ubuntuusers.de/topic/22098/ (German)

A temporary remedy is a "sudo laptop-mode start".

[Jeff Fortin Tam (kiddo)]

I am still here. Now, I'm running Dapper and Edgy, and this problem still is not gone. Any dev knows what actually could be going on here? If only I knew the cause, it would not be so irritating.

[Slade Winstone (slade-winstone-yahoo)]

I'm suffering from this too.

Running Edgy, 2.6.17-11-generic on an Asus A8M laptop. I'm seeing constant hdd access every 2-3 seconds.

Does anybody have any ideas?

[mimithebrain (mimithebrain)]

Same problem, no fix that I know of.

[padre999 (paimweb-business)]

The problem persists. Running an up to date Edgy Eft. Hdd access every 5 seconds with a loud click on my laptop and desktop. Very annoying. As I could see at Ubuntuforums it will be still there in Feisty.

http://ubuntuforums.org/showpost.php?p=2321063&postcount=45

Some activity or feedback from the developers would be appreciated!

[Laurent Bigonville (bigon)]

In my dmesg I have a line about commiting the fs journal:

kjournald starting. Commit interval 5 seconds.

Could be the source of the disk access?

[Jeff Schroeder (sejeff)]

Here is how you can find out the top offenders of what processes are using your disk:

First, you need to configure your system to log every single disk write. Here is how:
jeff@vishnu:~$ sudo -i
Password:
root@vishnu:~# /etc/init.d/sysklogd stop
root@vishnu:~# echo 1 > /proc/sys/vm/block_dump

Wait about 5 minutes and make sure that you hear the hard drive being accessed many times.

root@vishnu:~# dmesg | gawk '/(READ|WRITE|dirtied)/ {activity[$2]++} END {for (x in activity) print x, activity[x]}' | sort -rn -k2 | head -20 > diskaccess.log
root@vishnu:~# echo 0 > /proc/sys/vm/block_dump
root@vishnu:~# gedit diskaccess.log

Either paste the contents or attach diskaccess.log. That will tell you the top offenders that are writing to your disk.

[Jeff Schroeder (sejeff)]

Forgot to restart syslogd:

root@vishnu:~# /etc/init.d/sysklogd stop

Note that root@vishnu:~# is my root prompt and yours will be different.

[Jeff Fortin Tam (kiddo)]

I have tested on Feisty. This is the output if I stop syslogd beforehand. Wtf is going on with pdflush? I have 768mb of RAM and only 1/3 is in use, from what I understand that thing is there to flush processes to the hard drive? Oh yeah, and my computer was completely idle while doing the tests.

jeff@kaname:~$ sudo ./top_io.pl
Sleeping for 5 minutes to gather disk usage information...
nautilus(30456): Dirtied Inodes: 2
kjournald(3672): Writes: 28
dmesg(726): Dirtied Inodes: 18
cron(723): Dirtied Inodes: 5
epiphany(725): Dirtied Inodes: 2
epiphany(605): Dirtied Inodes: 2
kjournald(2209): Writes: 61
perl(726): Dirtied Inodes: 3
sh(723): Dirtied Inodes: 2
sh(724): Dirtied Inodes: 1
run-parts(724): Dirtied Inodes: 1
cron(722): Dirtied Inodes: 20
perl(588): Dirtied Inodes: 1
pdflush(137): Writes: 442

And this is the output if I don't stop it:
jeff@kaname:~$ sudo ./top_io.pl
Sleeping for 5 minutes to gather disk usage information...
dmesg(601): Dirtied Inodes: 18
dhclient3(596): Dirtied Inodes: 1
maxlifetime(594): Dirtied Inodes: 1
find(591): Dirtied Inodes: 1
dhclient-script(597):Dirtied Inodes: 1
avahi-autoipd(600): Dirtied Inodes: 2
sh(590): Dirtied Inodes: 2
sh(595): Dirtied Inodes: 1
syslogd(4304): Writes: 2
syslogd(4304): Dirtied Inodes: 175
run-parts(597): Dirtied Inodes: 1
gconfd-2(30408): Dirtied Inodes: 1
sh(592): Dirtied Inodes: 1
xargs(595): Dirtied Inodes: 8
dhclient-script(598):Dirtied Inodes: 1
call-dhclient-s(596):Dirtied Inodes: 2
cron(590): Dirtied Inodes: 5
perl(601): Reads: 1
perl(601): Dirtied Inodes: 3
pdflush(137): Writes: 384
pdflush(137): Dirtied Inodes: 47
run-parts(599): Dirtied Inodes: 1
dhclient-script(600):Dirtied Inodes: 1
kjournald(3672): Writes: 33
dhclient-script(596):Dirtied Inodes: 20
sed(594): Dirtied Inodes: 8
maxlifetime(592): Dirtied Inodes: 1
kjournald(2209): Writes: 635
cron(589): Dirtied Inodes: 20
sh(591): Dirtied Inodes: 1

[padre999 (paimweb-business)]

Thank you Jeff Schroeder.

I run the procedure you described on my laptop on an up to date Ubuntu Edgy Eft. Attached see the output of the two runs I did. Clearly on top of the list is kjournald. And there are two instances running of it!?

Anyway I am not so good in reading the log files, that's up to you experts.

[padre999 (paimweb-business)]

I support Jeff Fortin's suspicion that pdflush might be the culprit. In the first run there were not so many audible hdd clicks. In the second run I had almost constant hdd 5secs clicks. The number of pdflush writes in the second run were 61. That fits perfectly the number of accesses for a 5 minute period. 5mins=300secs. 300secs/5secs = 60

I have 512 MB RAM and 2/3 of the RAM are free.

[Peter Makowski (petermakowski)]

I have the same problem in Feisty. I have to switch to another distribution because Ubuntu with this bug is tooo slow. I just can't work.

[Sitsofe Wheeler (sitsofe)]

If you aren't running your machine in laptop-mode (which is an explicit setting) it is expected that the hard disk will be accessed every few seconds or so as data will be being written to the journal frequently to the partition journal keep things consistent in the case of a crash. Additionally, if you have atime reporting on (which is the default) then accessing a file (even to do a read) will cause a write as the access time of the file will be updated.

pdflush is a kernel thread and is almost certainly simply a messenger in all of this. To the best of my knowledge it just periodically flushes data back to disks.

[padre999 (paimweb-business)]

@sitsofe

the hdd is not accessed every few secs or so but in a precise interval accompanied with a clearly audible click. You can see this in the graphic I attached further up in the thread. I do not have this problem when I use other Linux Distros. I know this behavior only from Ubuntu and it is pretty nerve wrecking when you work on your computer in a silent environment.

[Peter Makowski (petermakowski)]

I noticed that disc also cannot be accesed for 4 seconds which causes lots of delays. I have to run applications twice sometimes and whole performance is poor. On the same computer Mandriva 2007 works perfect. Where''s the problem?

[Sitsofe Wheeler (sitsofe)]

padre999:
Are you using a laptop? Do you also have a CD/DVD drive in that laptop? If so this sounds similar to https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=213995 ...

[Jeff Fortin Tam (kiddo)]

I don't know for Piotr, but I, personally, suffered from this bug since ubuntu 4.10, on different laptops, with or without optical drives. My current one has the dvd/cdrw drive removed (because it's heavy and sometimes messes up suspend), so the redhat bug does not seem related for me.

[Peter Makowski (petermakowski)]

I'm not using a laptop, so... Any suggestions?

[Sitsofe Wheeler (sitsofe)]

Pirotr:
Nope - I think you've roundly debunked my suggestions : )

[Jeff Fortin Tam (kiddo)]

Hourra! Hourra! I succeeded! I was FINALLY able to make my laptop completely silent! And have its drive spin down (and stay in sleep) for the first time of my life! ^_^

What I did:
1- make the pdflush intervall MUCH longer. 10 minutes instead of 5 seconds (developers: could I get a comment on this? why isn't something more reasonable than 5 seconds the default?). I found the trick while googling: http://bbs.archlinux.org/viewtopic.php?pid=244453
2- disable access time on the partitions. I went into /etc/fstab and for each partition except the swap (that means / and /home for me), I added the option noatime (for example: "defaults,noatime")
3- I *think* that disabling klogd, sysklogd, cupsys and hplip (all of which can be disabled from "gksu services-admin") helps.

bonus level: if you want to test your hard drive sleeping after 5 seconds (extreme fast sleeping):
======================
sudo hdparm -S 1 /dev/sda
(note that the value of S = number * 5 seconds. So if I put -S12, it will be 1 minute before the drive spins down; replace sda by whatever your hard drive is)

Please give me your comments.

[Sitsofe Wheeler (sitsofe)]

Jeff:
1) By default (i.e. on a desktop) flushing every 30 seconds can reduce disk I/O pressure (i.e. it reduces pressure later on because that data is already on disk).
2) Was that not mentioned? Disabling atime is usually among the first suggestions...
3) I don't think I'd go so far as to disable klogd and sysklogd. If your computer crashes or something untoward happens you will have no evidence of what happened in the run up.

Just out of interest aren't a whole bunch of these things controlled when you use the laptop-mode package (see /etc/laptop-mode/laptop-mode.conf for all the settings that can be controlled when its installed) ?

[Jeff Fortin Tam (kiddo)]

well the fact is that disabling only access times, or only syslogkd, or whatever, yields no results. It's *only* when I disabled them all together that I was able to get some (drive) sleep!

Interestingly enough, the laptop-mode package is not installed on my machine. I would have thought that thing was installed per default if the computer was a laptop. I'll install it and see if it does anything different.

[Jeff Fortin Tam (kiddo)]

I'm not sure what laptop-mode does, but I just tested and it seems to me the pdflush trick is still needed. Of course nothing is certain :|

[Sitsofe Wheeler (sitsofe)]

Jeff:
You will need to reboot/start the initscript after installing it for there to be any affect. There is documentation on laptop-mode in /usr/share/doc/laptop-mode-tools/ ...

[Jeff Schroeder (sejeff)]

pdflush will write to disk less if you put this at the bottom of /etc/sysctl.conf:
# From powertop
vm.dirty_writeback_centisecs=1500

After that, run sudo sysctl -p for the new changes to take affect. This was a suggestion from powertop which does seem to lower disk writes by increasing the amount of time between disk writes.

[Jeff Schroeder (sejeff)]

Since bug 118303 has been resolved, there is another tool you can use. It is called blktrace and has a wrapper script called btrace that will spit out a (mostly) human readable format of what processes are eating your disk up.

Type mount to see what volumes / partitions you have mounted. Here is an example from my machine:
jeff@desktopmonster:~$ mount
/dev/sda5 on / type ext3 (rw,noatime,errors=remount-ro)
proc on /proc type proc (rw,noexec,nosuid,nodev)
/sys on /sys type sysfs (rw,noexec,nosuid,nodev)
varrun on /var/run type tmpfs (rw,noexec,nosuid,nodev,mode=0755)
varlock on /var/lock type tmpfs (rw,noexec,nosuid,nodev,mode=1777)
udev on /dev type tmpfs (rw,mode=0755)
devshm on /dev/shm type tmpfs (rw)
devpts on /dev/pts type devpts (rw,gid=5,mode=620)
lrm on /lib/modules/2.6.22-11-generic/volatile type tmpfs (rw)
/dev/sda1 on /boot type ext3 (rw,noatime)
/dev/sda7 on /home type ext3 (rw,noatime)
securityfs on /sys/kernel/security type securityfs (rw)
binfmt_misc on /proc/sys/fs/binfmt_misc type binfmt_misc (rw,noexec,nosuid,nodev)
debugfs on /sys/kernel/debug type debugfs (rw)

Since / is mounted on /dev/sda5, the actual device file is /dev/sda. Run 'sudo btrace /dev/sda' it will spit out a list (edited for brevity) of what processes are monopolizing your disk.
jeff@desktopmonster:/boot$ sudo btrace /dev/sda
[sudo] password for jeff:
  8,0 0 1 0.000000000 4788 Q W 2607452 + 8 [syslogd]
  8,0 0 2 0.000007077 4788 G W 2607452 + 8 [syslogd]
  8,0 0 3 0.000011407 4788 P R [syslogd]
  8,0 0 4 0.000013104 4788 I W 2607452 + 8 [syslogd]
  8,0 0 5 0.000036993 2640 U R [kjournald] 1
  8,0 0 6 0.000043375 2640 D W 2607452 + 8 [kjournald]
  8,0 0 7 0.000424356 0 C W 2607452 + 8 [0]
  8,0 0 8 0.000465454 2640 Q W 1036180 + 8 [kjournald]
  8,0 0 9 0.000467807 2640 G W 1036180 + 8 [kjournald]
  8,0 0 10 0.000469107 2640 P R [kjournald]
  8,0 0 11 0.000469898 2640 I W 1036180 + 8 [kjournald]
  8,0 0 12 0.000472288 2640 Q W 1036188 + 8 [kjournald]
  8,0 0 13 0.000474546 2640 M W 1036188 + 8 [kjournald]

Note that this is on a gutsy system. Try the btrace command and paste the output here.

[ubuntu_demon (ubuntu-demon)]

This might be one of the contributors to the famous Load_Cycle_Count bug :
https://bugs.launchpad.net/ubuntu/+source/acpi-support/+bug/17216
https://bugs.launchpad.net/ubuntu/+source/acpi-support/+bug/59695

[ubuntu_demon (ubuntu-demon)]

To use btrace you need to do this :

sudo aptitude install blktrace
sudo mount -t debugfs none_debug /sys/kernel/debug
sudo btrace /dev/sda

[ubuntu_demon (ubuntu-demon)]

btrace howto :
http://ubuntuforums.org/showpost.php?p=3722850&postcount=374

[Jim Qode (jimqode)]

I tried every trick on this list and nothing worked. I have a sony vaio VGN-N395E laptop with hitachi HTS541616J9SA00 disk.
Fedora does not do this. So there should be a way to make it work. I'm attaching btrace log.

[Dennis Murczak (dmurczak)]

Confirmed on my desktop with Kubuntu Gutsy. 20 or so seconds after the KDE desktop has appeared and there is still a lot of hard disk activity, it starts clicking like crazy, but without slowing down the system considerably, and after 2 or 3 minutes enters the well-known state where clicking occures every 5 seconds. After 15-20 minutes, clicking becomes infrequent and in most cases stops a few minutes later. Load cycle count of the disk is not exceptionally high (I checked that of course), but the seek and read error rates are insane.

I have a Hitachi hard disk too. As you might know, these are the successors to the infamous IBM "Deathstar" series which had a very bad disk-destroying clicking problem that could be fixed (mostly too late) by a firmware update. So I wonder whether Hitachi HD's need special treatment in terms of power management to not wear out too quickly.

I'm not affected on other distributions BTW and the ugly -B 254 fix doesn't do anything for me.

[Pneuma (nolan-lawson)]

I had the exact same problem with 7.10 Gutsy on a Dell Inspiron 1150 laptop - hard drive clicked every few seconds. The System Monitor I have in the panel confirmed this - with every click, there would be a spike in the graph. I managed to COMPLETELY FIX the problem based on what some people here (and elsewhere) have said, but I figured for the benefit of other Ubuntu newbies I would break down step-by-step what I did:

1) Enable laptop mode. This will ensure that the laptop-mode-tools program will actually be able to run. For some reason, Ubuntu has it automatically disabled. Go to Applications -> Accessories -> Terminal and type in "sudo gedit /etc/default/acpi-support", then your password when prompted. Search for where it says "ENABLE_LAPTOP_MODE" and set it from "ENABLE_LAPTOP_MODE=false" to "ENABLE_LAPTOP_MODE=true". Then save it and exit.

2) Get laptop-mode-tools and install it. Now, for some reason I couldn't find this program using the Add/Remove Programs feature, so I went to the main download page at <http://samwel.tk/laptop_mode/packages/debian> . (You have to download the debian package for some reason.) Download the second most recent one (1.34), and open it with the default program. When you open it up, it should alert you that a later package is available and you should download that one instead. Well, install this one anyway - just click "install." Within minutes after the installation, Ubuntu should give you a popup in the panel to let you know that a new version is available, so keep clicking through the windows to install that one. (This is just what worked for me. The most recent version from the web site gave me a warning about how I needed an “earlier” version of the package, so... whatever.)

3) Set laptop-mode-tools to run even when on AC Power, otherwise you'll still have clicking when you're plugged in. Go back to the Terminal and type "sudo gedit /etc/laptop-mode/laptop-mode.conf" and give the password if prompted. In gedit, search for "ON_AC". You should change "ENABLE_LAPTOP_MODE_ON_AC=0" to "ENABLE_LAPTOP_MODE_ON_AC=1" so that laptop-mode-tools will work even when your laptop is on AC power. Save and quit.

4) Turn on laptop-mode-tools. Go back to the Terminal and type "sudo /etc/init.d/laptop-mode start". This should start up laptop-mode. You can check to make sure that laptop-mode is on by typing "sudo /etc/init.d/laptop-mode status". A nonzero value after where it says “/proc/sys/vm/laptop_mode” indicates that it's activated. Now, in my case the "start" command reassured me that laptop-mode was on, but then the "status" command gave me a 0 value. That was because I was on AC power, and I hadn't done step 3. Yours should be a nonzero value.

5) The clicking sound should have stopped by now. (THANK GOD!) Now just set it so that laptop-mode will come on every time you boot. To do this, go to System -> Preferences -> Sessions, then click "Add" to add a new startup program. In the "Command" field, put "/etc/init.d/laptop-mode start". For "Name" or "Comment", put whatever you want.

This should do it! Enjoy the peace of mind that comes with not having your laptop screech at you eve...

[wangrui (cnsdqdwangrui)]

Hi, this problem had bothered me for a very long time one month ago. I used a flash disk to host the root partition to solve the problem. But after a month, I find the flash disk was easily lost data even with readonly all the time. So I moved back to hard disk. But the unionfs solution seems works well to solve this problem:

I'm using ubuntu AMD64, I found that /etc/mtab and /etc/hosts was access all the time. So if we mount etc into memory, the hard disk click will stopped immediately.

Put the following lines into the proper position of /etc/init.d/mountkernfs.sh:

/bin/mount -v -n --bind /etc /unionfs/ro/etc
/bin/mount -v -n -t unionfs -o dirs=/unionfs/mem/etc=rw:/unionfs/ro/etc=ro unionfs /etc

And you also need to put another script in rc0.d and rc6.d to save the changes in etc. And one in rcS.d to restore the changes. Otherwise all changes to etc will lost after a reboot.

[justanotherjoe (josephreger)]

I followed Pneuma comments from January 20, 2008, and viola - the hard drive clicking stopped, and has not returned. The only thing I did different was that I used the latest laptop-tools.

I have a T-21 IBM, running clean install of Ubuntu 7.1 with all of the latest updates to date. Absolutely was worried about this issue, and thank you for the perfectly clear instructions Pneuma!

[another_sam (anothersam)]

Just followed the steps posted by Pneuma
https://bugs.launchpad.net/ubuntu/+source/dbus/+bug/17878/comments/37
and they worked great.

I am seriously considering to buy an SSD drive to gain quietness on my laptop.

[Milan Bouchet-Valat (nalimilan)]

Well, this is nice to know, but you should not need to install laptop-mode-tools to avoid your HD being accessed everytime. There's still problem in some program.

[another_sam (anothersam)]

sh*t, what actually happened is that I was suffering the hdparm bug and my headache was the drilling sound of obsesive parking.

sudo hdparm -B 255 /dev/sda

http://ubuntuforums.org/showpost.php?p=5109109&postcount=9

(laptop-mode stills being a great app)

[christopherdstamper (christopherstamper)]

I'm having the same problem here, on an 8.10 laptop. I've actually had it since at least 8.04...

Any suggestions? I've noticed that most the posts here were for older versions of Ubuntu.

[justanotherjoe (josephreger)]

Hard disk parking has been an issue on my laptops (IBM T20 & T21) since at least 7.04 Beta. It appears to be the result of a lack of "open"-ness on APM standards, and succeeding arguments between manufacturers, programmers, and developers. Some argue that it may be shenanigans orchestrated in Redmond, with assistance from subjugated manufacturers. For 7.04 Beta & 7.10 I followed the advise in this forum. For 8.04 LTS I updated the firmware on my hard disk, then used the guidance at the Ubuntu support forum and the issue was resolved. (see http://ubuntuforums.org/archive/index.php/t-583913.html). If you want lengthy reading go to "http://brainstorm.ubuntu.com/idea/15153/", but I warn you that it sometimes strives towards Stonian conclusions (as in Oliver) utilizing Rumsfeldian logic.

[Milan Bouchet-Valat (nalimilan)]

justanotherjoe: This bug is not that about disk parking. Here we're dealing with disk access, which is obviously related, but different.

christopherdstamper: Enabling laptop-mode can solve your problem, in particular when on battery. Edit /etc/default/acpi-support and set ENABLE_LAPTOP_MODE=true, then enable disk management: set LM_BATT_HD_IDLE_TIMEOUT_SECONDS=30 (or similar value, in seconds) and CONTROL_HD_POWERMGMT=1. Then restart.

This bug is strange, can come from different issues, both software- and hardware-related. Are there other people still seeing this *precise* every-3-second access problem?

[christopherdstamper (christopherstamper)]

@Milan: Are there any downsides to using laptop mode?

For me, killing klogd fixed the problem. Guess I don't need the kernel log that much anyway...

[Milan Bouchet-Valat (nalimilan)]

Normally laptop-mode should not raise any problems, if you configure it so that your disk heads don't park too often. If klogd was your problem, can you have a look at your log files to find the precise cause? It would be nice to fix this. For example, use the Log Viewer (System->Administration) to see what log messages appear and make your disk wake up. NetworkManager has such a bug in some cases for example. Thanks!

[christopherdstamper (christopherstamper)]

@Milan: There are no applicable entries in any of the logs, or in dmesg. Apparently, it's just accessing the hd, and doing nothing.

 Gotta go kill klog before it makes me crazy... ;-)

[Milan Bouchet-Valat (nalimilan)]

If you have some time, it would be useful to test whether enabling laptop-mode solves the problem, and else report a bug against the sysklogd package. Getting a better disk parking management in Ubuntu would be really great.

[christopherdstamper (christopherstamper)]

@Milan: Enabled laptop mode, rebooted. Disk access continues, though not as predictable. BUT, now I don't hear the annoying 'disk-unpark' noise, or whatever you want to call it. So I guess it's a partial fix...

[Milan Bouchet-Valat (nalimilan)]

This means that you've set the HD_IDLE_TIMEOUT_SECONDS to a value high enough to prevent the disk from sleeping at all. I'd have hoped the opposite. If you don't hear the disk heads unparking, then laptop mode doesn't fix anything but has just the side-effect of keeping your disk up. Please report a bug against sysklogd, it must be the culprit.

[ne (office-sico)]

I had this strange behaviour too.

On a MB (GA/nforce560) I used a Seagate 7200.11 (SATA) and a fresh install of Ubuntu 8.10.
I noticed a HDD activity every 4-5 sec, which was very annoying.
Because of the Firmware bug of the Harddisc I decided to replace the HDD with another one.
I took a Seagate 7200.10 (IDE) and never noticed this problem again.
-> so the problem seems to be caused by something SATA-related (?)

But it became more crazy: After flashing the firmware of the 7200.11 I put it into another
PC (GA/nforce3-150) and also did a fresh install of 8.10. The problem did not occur on that system!

So in my thinking the cause of this problem is:
- either the firmware of the hdd (cache flush?)
- the driver or some driver related software of the nforce560 sata controller

I hope this can help you locating this effect.

[Milan Bouchet-Valat (nalimilan)]

If it has something to do with the HD's firmware, I don't expect that somebody will fix that. Thanks for your informations. Though, please note that there have been HD-related upgrades recently in Intrepid, so maybe when you installed Ubuntu on the other computer, the fix came in, and this is why you don't see the problem anymore. This is just a possible explanation, maybe not the right one.

For now I'll close the bug because you don't experience it anymore (whatever the cause of the fix), and I don't either. So we can't debug it! Moreover, it's reported against d-bus, which does not seem to be the culprit here.