tar xvfz hangs the machine

[Chakravarthi]

Hi,

I have installed Redhat 8 on pentium 4 DELL machine, when i try to untar a BIG file machine hangs,
i have observed that, when the tar command is running, kjournald takes more CPU and after few seconds get a kernel panic message,

can anyone explain me what is this kjournald, why this is used?

and how to take care of this kernel panic problem

thanks
chakri

[Mark Grant]

kjounald is your ext3 journaling daemon. You might consider changing the way you journal or even changing your journalling file system.

Check out this article for changing ext3 journalling modes.

http://www.linuxplanet.com/linuxplanet/reports/4136/5/

[Chakravarthi]

if i look into the var/log/messages
i have the followin error message

Kernel: mtrr: no more MTRRs available

[Mark Grant]

This seems to come up a few times.

There is a document in the kernel sources if you have them. /Documentation/mtrr.txt which helps to explain the mtrr stuff. Failing that, you can look here

http://www.iglu.org.il/lxr/source/Documentation/mtrr.txt

I don't think this is going to be the reason for your problem though.

[Alexander Chuzhoy]

Look here for help:
http://www.uwsg.iu.edu/hypermail/linux/kernel/0301.3/1029.html

[Chakravarthi]

I'm not sure what is the problem,, kjournald and gdmgreeter takes more cpu and hangs the machine when ever there is more disk activity

[Mark Grant]

Try untarring to a non-ext32 filesystem and see if that helps. IT might be worth posting the kernel panic you are getting. Also, let us know which kernel version you are using.

[Chakravarthi]

i tried on ext2 partition still the same behavior

i'm using redhat 8 with 2.4.18-14 kernel

thanks
chakri

[Fred Ruffet]

How "big" is the tar ?

When you use an ext2, you can't have the same symptoms (kjounald not used by ext2). What's happening exactly when running on ext2 ?

[Martin P.J. Zinser · Accepted solution]

Hello Chakri,

can you to a tar -tzvf to list the archive in
question. How big are the file sizes reported?

Greetings, Martin

[Kiyoshi Miyake]

Hi,

Try the following commands, in order to carve where a problem is.
(file read, write, gzip, tar...)

1) cat BIGfile > /dev/null
2) gzip -dc BIGfile > /dev/null
3) cp BIGfile BIGfile.new
4) gzip -d BIGfile
5) (after 4.) tar tvf BIGfile.tar(?)
6) (after 4.) tar xvf BIGfile.tar(?)

Thanks.

[Stuart Browne]

and how big is 'Bigfile'.

[U.SivaKumar_2]

Hi,

Ensure that you have GNU tar 1.13 and not less than that.

#rpm -qa | grep tar

Ensure that you have gzip version 1.3.3 or above.

#gzip -V

regards,

U.SivaKumar

[dirk dierickx]

if it happens on ext2 you can be sure it had nothing to do with ext3.

be sure to get the latest patches for rh8 (latest kernel release etc).

it could still be a hardware problem, can you post the panic messages here?

try running a memory checker/stress program like mem86: http://www.memtest86.com/

[Fred Ruffet]

> I'm not sure what is the problem,,
> kjournald and gdmgreeter takes more cpu
> and hangs the machine when ever there is
> more disk activity

A while ago, I made tests on different journalized FS for Linux. EXT3 and Reiserfs seems to generate a high CPU activity, even in normal use. Maybe you have file size, or file count that make this default evident.

I would suggest you to take a look at XFS. It's a really good journalized FS (ported from Irix by SGI themselves). I never had any problem with it (made crash-tests and performance tests). I use it in combination with LVM.

[Chakravarthi]

it 600MB file

[Mark Grant]

This might be a bit unlikely but I have seen similar issues when the CDROM drive was set as a master and was in use at the time. Probably not in this case though.

If you could post the kernel panic it might help diagnose the problem.

[Martin P.J. Zinser]

Hello Chakri,

Thank you, the 600 MB is an important information as it allows to execlude problems with tools hitting the 2 GB limit (as I and obviously some other posters had suspected).

After some research I also do not expect that the mtrrs are really the problem here, although I certainly would not hurt if you could post the output of a cat /proc/mtrr here.

Else how about space in the file systems, i.e. not only the target filesystem, but also things like /tmp or /var.

Greetings, Martin

[Chakravarthi]

/var and /tmp have enough space,, 20% used

[Stuart Browne]

Ok. No real explanation then.

Best guess would be a controller-driver issue. As you are using DELL machines, I don't know really where to begin to tell you to look, as I've never used one for Linux.

The Perc series of RAID controllers will be using either the 'aacraid' or the 'megaraid' drivers. Both of these have had significant updates since the RH8, and it might be worth seeing if you can update your appropriate driver.