View previous topic :: View next topic |
Author |
Message |
Major Apprentice
Joined: 27 Feb 2004 Posts: 198 Location: Sherbrooke, Qc, Canada
|
Posted: Sun Aug 14, 2005 8:28 pm Post subject: kjournal taking CPU usage |
|
|
As soon as I have some HD activity kjournal kicks in and take most of the CPU same for pdflush.
What are those two process (Kernel deamon as far as I can guess)
But why are they writting all the time?
Code: | # uname -a
Linux tux 2.6.11.11 #6 Fri Aug 12 08:14:36 EDT 2005 i686 AMD Athlon(tm) XP 2600+ AuthenticAMD GNU/Linux
|
I have a SATA HD... _________________ Major over and out |
|
Back to top |
|
|
widan Veteran
Joined: 07 Jun 2005 Posts: 1512 Location: Paris, France
|
Posted: Sun Aug 14, 2005 10:37 pm Post subject: |
|
|
pdflush is the kernel's dirty page writer. It writes modified memory pages that belong to open files back to disk. kjournald is ext3's journalling daemon. It writes the filesystem metadata to the disk. So it's normal that both of these become active when there is disk (actually filesystem) activity. |
|
Back to top |
|
|
Major Apprentice
Joined: 27 Feb 2004 Posts: 198 Location: Sherbrooke, Qc, Canada
|
Posted: Sun Aug 14, 2005 11:07 pm Post subject: |
|
|
ok but not as they do right now ... they are using like 40 % each...
This is what dmesg ends up after coping a 300 MB file from one HD to an other
Completly filled with that kind of message....
Code: | # dmesg | tail
reiserfs/0(382): WRITE block 262904 on sda4
reiserfs/0(382): WRITE block 262928 on sda4
reiserfs/0(382): WRITE block 6817736 on sda4
reiserfs/0(382): WRITE block 19136512 on sda4
reiserfs/0(382): WRITE block 22282344 on sda4
reiserfs/0(382): WRITE block 27525256 on sda4
reiserfs/0(382): WRITE block 28049424 on sda4
bash(2907): READ block 7079288 on sda3
bash(2907): READ block 30933448 on sda3
bash(4873): READ block 27337184 on sda3
|
Just copying files slow the PC to a none usable point ... mice laggs time slows down (gkreallm) just unusable ... _________________ Major over and out |
|
Back to top |
|
|
widan Veteran
Joined: 07 Jun 2005 Posts: 1512 Location: Paris, France
|
Posted: Sun Aug 14, 2005 11:36 pm Post subject: |
|
|
Those messages are only logged (by the block layer) if you have /proc/sys/vm/block_dump set to 1. Normally it should be disabled by default. To disable it, run (as root):
Code: | echo 0 > /proc/sys/vm/block_dump |
Try it and see if that helps. If it does, you will have to find why block_dump is enabled to begin with (this feature exists so that people are able to track down processes that do I/O, when they are causing unexpected disk spin-ups on laptops).
I also hope you're not logging KERN_DEBUG messages with syslog... as the logging will cause disk activity, which will itself cause more messages, which causes more logging... and so on, and the machine can slow down very easily if you do. |
|
Back to top |
|
|
Major Apprentice
Joined: 27 Feb 2004 Posts: 198 Location: Sherbrooke, Qc, Canada
|
Posted: Fri Aug 19, 2005 8:29 pm Post subject: |
|
|
The issue was that DMA was not set for that HD ...
If not it solve the matter _________________ Major over and out |
|
Back to top |
|
|
|