background disk write problem

[vc]

Problem description:  background disk writes to the / partition are occurring constantly.  This behaviour is noted on all Minime 09 installs.

Test system:  an older Acer TravelMate laptop, equipped with a new 80GB IDE hard drive.

Installation details:  a fresh install of Minime 09, installed with standard default settings except for GRUB menu entry amended to include only 'acpi=on nosmp' parameters; 'resume=' and 'splash=silent' parameters were removed.  Install has not been updated at all, and only three additional packages were added from the KDE3 repo through Synaptic (autologin; gkrellm; ksnapshot.).

Partitioning details:

/dev/hda1 = swap
/dev/hda2 = /
/dev/hda5 = /opt
/dev/hda6 = /tmp
/dev/hda7 = /usr
/dev/hda8 = /var
/dev/hda9 = /home

All partitions except swap are formatted ext2, with default /etc/fstab mount options.

Test conditions:  fresh boot, with eth0 disconnected, and Update Notifier manually exited.  GKrellM started first; system left entirely alone afterwards, for a period of about forty minutes; then, KSnapshot started and screencap obtained (copy attached, below.).

Referring to the screencap image, a clear pattern of disk accesses over a period of time has been recorded.  With the exception of the KSnapshot startup activity burst, all partitions except /dev/hda2 show no activity - /dev/hda2, however, shows a regular and constant pattern of disk writes of 8.2KB in size, occurring approximately every ten seconds.

What is causing these constant background disk writes?  Are they due to some system monitoring daemon, or could they perhaps be due instead to kernel message logging?  How may they be eliminated, please?  Any advice or assistances would be greatly appreciated.

[attachment deleted by admin]

[vc]

Update - I have now installed one additional package, KSystemlog, and have attached its output as a textfile.  Unfortunately, however, this output does not seem to indicate whatever may be causing the pattern of constant disk writes.

To explain a bit further, my concern regarding these disk writes is that they are very problematic for any installs performed on a system which depends upon SSDs for storage, particularly in cases where the system SSD is comprised of flashram chips that are already several years old and physically non-replaceable (as with, for example, the original EeePC 700/701 series.).  Assuming an older flashram lifetime rating of, say, one million write cycles, would therefore mean that with this particular pattern of constant disk writes occurring at ten-second intervals, a flashram failure would occur in only 116 days.

[attachment deleted by admin]

[muungwana]

try to disable system logging and see if they still happen. You can do that by starting "pcc -> system -> manage system services" and uncheck and stop "syslog" service. Do they still happen?

[vc]

try to disable system logging and see if they still happen. You can do that by starting "pcc -> system -> manage system services" and uncheck and stop "syslog" service. Do they still happen?

Hmm... the results obtained so far seem intriguing, muungwana.  I began by performing the action you suggested; however, as that did not seem to have any effect I then proceeded to stop other system services, one by one, yet again with each having no effect upon the /dev/hda2 disk writes.

Ultimately, every service except dm had been stopped - I then left the machine undisturbed for a while, and when I returned I took the following attached screencap:

[attachment deleted by admin]

[vc]

Aha!  The attached screencap now shows system behaviour with only the messagebus service disabled.  Does dbus have a surreptitious logging function?  If so; does it also have an associated config file which may be edited?

[attachment deleted by admin]

[wedgetail]

vc
Currently I am not very active on forum, only visit infrequently due to heavy workload  out doors.  This problem I believe may not be only on MiniMe09, but I am not completely sure because I do have a running MiniMe copy I use sometimes for fun.

I believe the subject of constant or very frequent read/writes have come up a few times on Forum. I investigated this after installing a new hard disk which was not low noise type and I found/find the constant clonking noise concerning.  I think I found it was a problem that I had to live with.

I am not in luck searching the forum and getting good results tonight, but I have found one thread happening to mention similar and with pointers to what was observed

http://www.pclinuxos.com/forum/index.php/topic,64928.0.html

I will take note as I move from system to system and see where it is I have the problem.  Actually I notice that the PCLOS2009 I am using right now has very frequent hard disk access but the drive is low noise type and only the hdu LED and gkrellm shows the activity. 

[vc]

Thank you, wedgeling!  That is indeed useful information; I shall pm the member accordingly in order to hopefully learn some more of the details.

[wedgetail]

Most annoying, but so far I have not been able to find my own research on this topic of unusual drive activity. Advanced years I guess is my problem.

One thing, did you try "top" in terminal, though I don't remember if that is available in MiniMe.

KSystemlog I would not rely on as showing all messages as I have tried using this to identify information during boot process. I may have used it wrongly, in that I assumed that KSystemlog would record what I was after.

Any chance that it would be some SMART activity from your hard drive, hopefully "top" would show if SMART running. (I guess that the SMART tools are not installed default in MiniMe, so unlikely culprit)

These comments are made with a fair chance of errors so be careful. I can't check as I am on a windows 2000 pc.

[vc]

Unfortunately wedgeling, Harold's suggestion did not seem to produce the desired result; alas, the constant disk writes to / continue unabated.  I am therefore still wondering if this problem may indeed be more dbus-related, as disabling that particular service seems the only means so far of eliminating this undesired disk write behaviour.

[wedgetail]

I have just started up my LDXE and checked that it does indeed have 'top'.  I found however that I must have been away from computing experiments for longer than I realised as there was a lot of updates, which I have set going for about an hour. I really should be out in the sunshine working the garden but I am rather curious so I will hang in here for a bit, on and off.

So when you are close to having shut all services, then what can you gather from 'top' (run from terminal)

Searching the web about DBUS I found: Dbus is not a KDE3 implemented, well that was my interpretation.  In KDE3 it is DCOP and in KDE4 an 'improved DCOP' implemented, named Dbus?  (That is about the extent of my knowledge, I understand the concept, I think but the practical side is above my head)

Do you mean DCOP ?

[vc]

Well, wedgeling, I'm not certain - however; if you would care to read through the foregoing messages, please look at the last screencap image I posted, in the message containing this statement:  "The attached screencap now shows system behaviour with only the messagebus service disabled."

The messagebus service was the only service I'd disabled then; all other services were still running at that time, yet the undesired write activity had ceased altogether.  The description shown in the PCLinuxOS Control Centre for the messagebus service refers only to dbus; therefore, I had assumed that the problem could likely be dbus-related.

I do not wish to shut down dbus entirely, of course, for rather a number of functions depend upon it - and, there is the consequent matter of both CPU usage and process loading instantly rising to their maximum values following such action as well; that behaviour is also quite undesirable.  Therefore, I asked if there was perhaps a dbus logging function occurring in the background, as it may then instead be possible to eliminate the disk write behaviour by simply disabling only that logging function.

Or, I could be following the wrong direction entirely?  Again, that is why I am not certain.

[wedgetail]

vc
I have read the message you refer to and you talk about 'dm'  which is a service, so I went looking around the 'dm' unfortunately I jumped to the conclusion that you had made this short for 'dmesg', I was scratching my head over this.  While in services I even looked at the 'messagebus' but absentminded, not in 'info' as I was starting to worry about updating, thinking this may be why I did not see what you had ......... As you can see I totally lost my clear thinking.

Anyway I did the updating and the system seems to be fine despite a number of ' .. could not find such a file' which I don't have time to follow up on and LXDE seems to run fine.  

After your last message I went again to 'services' in the updated LXDE found 'messagebus' and looked at the info and here is the mention of DBus, well I guess this must mean that DBus is implemented in KDE3.

Yes I can follow your thinking perfectly now.  If DBus has a finger in every pie it gets hard to figure which one.

What about looking at 'top' records with and without messagebus active? Can you find something that differs? If it happens regularly every second or better then the process ought to percolate to the top o the list?

[vc]

I have attached below an archive (in .rar format - please rename accordingly, due to upload difficulty), which contains some pertinent files.  The 'top1.txt' file was created upon a fresh reboot, and prior to stopping the 'messagebus' service, whereas the 'top2.txt' file was created about ten minutes after stopping the 'messagebus' service; likewise also for the 'ps_axm-FL_1' and '2' .txt files, which were created to provide further detail by running:

ps axm -FL

in a root terminal.

The 'dm' service I had referred to in a previous posting could likely be a 'desktop manager' service, for when that particular service is stopped, KDE immediately disappears and one is left with nothing but a commandline prompt.

[attachment deleted by admin]

[wedgetail]

vc
I have had a look at the 'top1 and top2' files. I don't see anything that may give a clue.

I have had one quick look at the two 'ps" files, they are a bit too complicated for me in that I recognise the names but are far too hazy about what they do in detail. I did notice that there was reference to both 'dcopserver' and 'dbus'. I did not understand if it meant that 'dcop' was stopped and 'dbus' took over.

'dm' according to the INFO is "Launch Graphical Display Manager", so no surprise that you have better keep that open.

Right now I think I am out of further ideas. I am using pclos2009 KDE3.5 and the hard disk is happily clunking away a lot of the time, further gkrellm report a mark every 30 seconds when I just let the system run.

However I have not made any attempt to shut down anything.

[vc]

Hmm.  An interesting observation:  apparently, the mysterious background disk write activity on my wife's EeePC has now ceased.  This was the case that had caused my initial concern, and was therefore the reason for this thread.  Some months ago, the original Minime 08 install on her 701 suddenly ceased functioning; investigation revealed the problem was caused by the failure of its SDHC card, upon which the /tmp, /var/, and /home directories had been placed.  I had then replaced the card, updated the machine's BIOS, and re-installed with Minime 09 - then discovered the background disk write problem.

I was quite concerned, yet didn't know how to rectify the mysterious behaviour - thus prompting me to begin this thread.  I subsequently confirmed the behaviour in other installs since, and have not found any solution - yet now, on that one machine at least, the problem seems to have somehow resolved itself?  How could that be possible?  Her usage pattern is very minimal; she does use her EeePC often, although basically only for browsing her two favourite websites:  YouTube and FaceBook.

Meanwhile, the background disk write behaviour yet continues on my other installs - such as the 2010 b2 install on my recently-acquired Acer TravelMate 2200, which is now serving as my 'main' computer these days.  One additional difference was noted, though:  with the Minime 09 install on my wife's EeePC, the background disk write behaviour was apparent immediately - whereas with the 2010 b2 install on my TravelMate, it was not initially apparent and instead began to develop only after several days of usage.

Another possibly pertinent observation is that the behaviour also seems to be dependent upon the selection of desktop environment.  With the TravelMate setup I had also installed some alternative windowing managers - and while the disk write behaviour does occur with KDE, I have noticed that the same does not occur when I boot the machine into the Enlightenment environment.  I shall repeat the test with some other window managers now, and report the results presently.