Ocfs users - Jan 2005 - Q: Copy/Move of a to OCFS file system causing system hang

Greetings,
We were looking for some advise on an OCFS hang: Out of memory error.
We are on red hat version 3 , kernel 2.4.21-27.0.1.ELsmp  and ocfs  1.0.13-1.
 
If we  try to move/copy a 5GB compressed file into a ocfs file system,  the 
command gets killed by kernel after 70% of work with the following log in  
/var/og/messages:
 
Jan 31 16:02:41 conrep-d04 kernel: 4177919 pages of RAM
Jan 31 16:02:41  conrep-d04 kernel: 3817463 pages of HIGHMEM
Jan 31 16:02:41 conrep-d04  kernel: 205990 reserved pages
Jan 31 16:02:41 conrep-d04 kernel: 1146612  pages shared
Jan 31 16:02:41 conrep-d04 kernel: 0 pages swap cached
Jan 31  16:02:41 conrep-d04 kernel: Out of Memory: Killed process 3595 (bash).
Jan 31  16:02:41 conrep-d04 kernel: Out of Memory: Killed process 3595 (bash).
Jan 31  16:02:41 conrep-d04 kernel: Fixed up OOM kill of mm-less task
Jan 31 16:17:04  conrep-d04 syslogd 1.4.1: restart.
 
 
This process has worked for us in development. We have copied 100GB of  files 
in development.
 We are transitioning to production to new HP Linux servers. However,  we are 
stuck with this error. We are waiting on Oracle support tar  4282257.999.
 
Meanwhile, Any advise on what to look for the memory management error. The  
system is idle. the cp or mv is the only command.
 
OCFS details
ocfs partition/fs size 1 TB
number of ocfs partitions/node 25
number of nodes 8
block size 128
 
Is kernel kernel .2.4.21-27.0.1.ELsmp  an issue? Development is on  
2.4.21-27.0.2ELSMP.
 
 
thanks
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
http://oss.oracle.com/pipermail/ocfs-users/attachments/20050131/0881d296/attachment.html

you should use cp or mv with o_direct, use the coreutils on our website,
if you are not already doing so.

I would assume the prod server has more ram than the dev one ? 

On Mon, Jan 31, 2005 at 09:15:46PM -0500, Ramgrid@aol.com
wrote:
> Greetings,
> We were looking for some advise on an OCFS hang: Out of memory error.
> We are on red hat version 3 , kernel 2.4.21-27.0.1.ELsmp  and ocfs 
1.0.13-1.
>  
> If we  try to move/copy a 5GB compressed file into a ocfs file system,  the
> command gets killed by kernel after 70% of work with the following log in  
> /var/og/messages:
>  
> Jan 31 16:02:41 conrep-d04 kernel: 4177919 pages of RAM
> Jan 31 16:02:41  conrep-d04 kernel: 3817463 pages of HIGHMEM
> Jan 31 16:02:41 conrep-d04  kernel: 205990 reserved pages
> Jan 31 16:02:41 conrep-d04 kernel: 1146612  pages shared
> Jan 31 16:02:41 conrep-d04 kernel: 0 pages swap cached
> Jan 31  16:02:41 conrep-d04 kernel: Out of Memory: Killed process 3595
(bash).
> Jan 31  16:02:41 conrep-d04 kernel: Out of Memory: Killed process 3595
(bash).
> Jan 31  16:02:41 conrep-d04 kernel: Fixed up OOM kill of mm-less task
> Jan 31 16:17:04  conrep-d04 syslogd 1.4.1: restart.
>  
>  
> This process has worked for us in development. We have copied 100GB of 
files
> in development.
>  We are transitioning to production to new HP Linux servers. However,  we
are
> stuck with this error. We are waiting on Oracle support tar  4282257.999.
>  
> Meanwhile, Any advise on what to look for the memory management error. The
> system is idle. the cp or mv is the only command.
>  
> OCFS details
> ocfs partition/fs size 1 TB
> number of ocfs partitions/node 25
> number of nodes 8
> block size 128
>  
> Is kernel kernel .2.4.21-27.0.1.ELsmp  an issue? Development is on  
> 2.4.21-27.0.2ELSMP.
>  
>  
> thanks
> _______________________________________________
> Ocfs-users mailing list
> Ocfs-users@oss.oracle.com
> http://oss.oracle.com/mailman/listinfo/ocfs-users

We have a need to copy files from a remote machine to
an OCFS partition.  Are there any plans to provide a
secure copy (scp) or remote copy (rcp) utility with
the o_direct flags?  Or maybe they exist but we just
never installed them.


		
__________________________________ 
Do you Yahoo!? 
Yahoo! Mail - Find what you need with new enhanced search.
http://info.mail.yahoo.com/mail_250

On Wed, 2005-02-02 at 13:35 -0800, David McWhinnie
wrote:
> We have a need to copy files from a remote machine to
> an OCFS partition.  Are there any plans to provide a
> secure copy (scp) or remote copy (rcp) utility with
> the o_direct flags?  Or maybe they exist but we just
> never installed them.
I had found these for Red Hat AS 2.1 while researching ocfs recently..

http://oss.oracle.com/projects/coreutils/files/AS2.1/i386/

"openssh-clients-3.1p1-14.1.i386.rpm   O_DIRECT enabled version of scp
(--o_direct) also requires updated fileutils with O_DIRECT enabled cp"

Not sure if that helps, maybe there is other versions up there as well.

Bruce Holzrichter