For some reason, when I attempt to copy a file, using cp, from one location to another the cmd is stuck (or taking an unusually long time).

Usually copying a 1 GB file will take 3-4 mins, so far it's at 11 mins and yesterday was stuck for several hours.

Is there a way I can see WHY or WHAT is going on?

Thanks

The AIX screen:

rm: Removing /oracledata/IFSP/ifsapp_archive_data01.dbf
rm -fe /oracledata/IFSP/control01.ctl
rm -fe /oracledata/IFSP/control03.ctl
cp -p /ifsdoc/Backup/BackupofIFSL/oracledata/IFSL/ifsapp_archive_data01.dbf /oracledata/IFSP/ifsapp_archive_data01.dbf

The file is about 1GB.

I did a test.

I FTP'ed a 6MB file to /oracledata/IFSP and it took about 15-20 seconds. Copying the same file to any other filesystem on the same server took a second.

I also noticed that the file I was copying (the 1GB file) was actually copying, very slowly. 53MB in 30 minutes.

Is there some chkdsk kind of thing I can perform. Note, /oracledata also houses the live, IFSL folder which is "in use".

Thanks

I think this is the problem:

/dev/redo1_lv -- /redolog1 jfs2 32768000 rw yes
no
/dev/redo2_lv -- /redolog2 jfs2 32768000 rw yes
no
/dev/archive_lv -- /archivelogs jfs 65536000 rw yes
no
/dev/oradata_lv -- /oracledata jfs2 405536768 cio,rw yes
no
/dev/oraindex_lv -- /oracleindex jfs2 305659904 rw ye
s no
/dev/ifsdoc_lv -- /ifsdoc jfs 403701760 rw yes
no


The CIO is known to cause performance issues when copying at OS level.

We are not sure how it got set on, but how can we turn it off!!!

You can disable this Concurrent IO option by using smitty jfs2, go to change /show ... locate your file systems and remove the cio from Mount options. You can also edit the /etc/filesystems by hand to remove this option from the stanza in question. Remount of /oracledata is required for changes to take effect.

The only problem here is Oracle performance. By using cio option on the mountpoint the databases performance is greatly increased. If you choose to change this configuration you migth want to look in to tuning vmo parameters to decrease the degree of file system buffers used by the system.

Thank you Spell - though it was switched on accidentally (or mysteriously).

We were running without it since the last time we had it ON, with JFS2, we lost a filesystem totally - which was not a great position to be in!

The os commands "cp" for example run incredibly slow (unacceptably so) although perhaps Oracle is faster. (Can't tell at the moment - since users only moan when it's slow, not parise when it's fast!)

I will unmount, change, remount I think!

Cheers