Remote Copy

[Nobody's Hero]

We are going to resize some oracle DB's on a prod system. So I created equal filesystems on another dev box so we can copy the DB files to another system for fall back if something goes wrong. Our plan is to use rcp to copy the files to the dev system. Question, would I have better performance using rcp or should I nfs these mount points to the prod system and just use copy?
10x
RPM

[Robert-Jan Goossens · Accepted solution]

Hi Robert,

rcp or a combination of find and cpio are fine.

# find /dir | cpio -ov | remsh server " cd /dir ; cpio -idvum "

# rcp -p files server:/dir/

Hope it helps,

Robert-Jan.

[Pete Randall]

Robert,

I'm not sure but I would think that you would be better off to avoid the overhead and complications of NFS.


Pete

[Mark Grant]

Actually, I think ftp might be marginally faster but definately I think NFS is not the optimum choice. rcp is fine.

[Alzhy]

Or use vxdump/vxrestore:
.
On Source server:
.
vxdump 0f - /ora001|remsh tgtserver "(cd /ora001;vxrestore rf -)"
.
Of course you must set up tgtserver to allow srcserver to remsh to it.


HTH.

[Alzhy]

And I would like to add: trust the tools that come with your Filesystem. In your case, I suppose it is VxFS (aka OJFS) so it is just fitting and proper that you use the tools native to it -- namely vxdump and vxrestore. If you update to VxFS/OJFS 3.5 (free btw), you get the ability to do vxdump/vxrestore on directories and files -- instead of whole Filesystems in the earlier versions.

[Geoff Wild]

We do this a lot - best performance we found was to do something like:

/usr/local/bin/tar -cf - /data | compress -c > data.tar.Z

Then rcp the file across the WAN...

That tar is the gnu tar...

RGds...Geoff