Copy large FS between servers.

[Fredric Persson]

Hi!
I having problem with growing file size when using rcp between 2 servers. We have 2 identical servers (N4000) with same size of the File Systems and block etc. Some of the source FS is almost full (2MB of 4GB free), but when using rcp -rp : the target FS is getting full. I know that this can it happends and why, but does anyone know how to fix the problem(the best way of copy large file between servers)? Tested with cpio but hard to copy it remotely if you don?t NFS mount the target FS. (It is 45 SAP database File Systems (Oracle) ? 4GB each). Used Omniback backup - restore, but it must be other ways. (I know that a SAN would be nice - Business Copy).

Thanks!
Regards,
Fredric

[Bill McNAMARA_1]

Try tar..

Later,
Bill

[Fredric Persson]

Bill,
do you mean that I sould tar every FS before copy it? I would need extra space to perform a tar and I what to copy everything in one command, like rcp -rp /oracle//sapdata* :/oracle/.

//Fredric

[Michael Tully]

Hi,

Would it be possible to set up an automatic ftp job that uses .netrc
Is job the sort of thing that would be done on a regular basis?

Business copy would be nice if you have the affordable space!

More information on how your two systems are setup may be helpful.

-Michael

[Alexander M. Ermes]

Hi there.
Other possibilities :
- XWindow ftp like in ExCeed, which can do server to server transfer
- try to write a nice ftp script to ftp these files
- tape utilities ( cpio, tar, fbackup )
cpio script attached

Rgds
Alexander M. Ermes

[Tony Constantine_1]

On server ServerA

cd
tar cvf - . | remsh ServerB 'cd /; tar xvf -'

[Fredric Persson]

Thanks all!
I think I will try out the "tar solution" and see if it works.

//Fredric

[Volker Borowski]

Fredric,

having only 2MB of 4GB free on a sapdata filesystem is something you never should have.

brrestore requires 2% freespace for restore in SAP environment otherwise it will not work (Do not know if there is a patch on sapserv yet). It will simply complain "no free space" and quit!

See if you can shrink the datafiles from 2000M to 1950M or so. This should bring you to 97%-98% usage. Since you copy the system anyway, may be you could try "brrestore -q" at the target, to see if it complains.

As for the copy action, there is a nice option "disk_copy" for brbackup now. Only problem is that NFS is slow, but you can use several copy-processes in parallel!

Do not know if this helps.
Volker

[Fredric Persson]

Hi!

The FS are filled up with oracle files (Table Spaces) and they are not 100% used. But you have a point that it can cause problem.

Thanks!
//Fredric

[Roger Baptiste]

I use this for large copies:

******
find | cpio -ov | compress - | remsh '(cd /; uncompress - | cpio -ivmud)'
****

The src_dir and dest_dir
parameter needs to be changed
as per your requirement.

-raj

[Thierry Poels_1]

Hi,
I strongly agree with Rayman. If you have a 10 or 100 Mbit network, then your network will probably be the bottleneck.
Go for tar/cpio --> compress/gzip --> remsh otherserver "cd /destination; uncompress/gunzip --> tar/cpio"

regards,
Thierry.

[Tommy Palo]

Take a look at "rdist" at:
http://gatekeep.cs.utah.edu/hppd/hpux/Networking/Admin/rdist-6.1.5/

[Magdi KAMAL]

Hi Fredric,

If you have the choice between tar and OmniBack, I would really prefer to use Omniback because of it's speed ( Streaming mode ) specially for huge file ( Your case ).

In my environment, mixing OmniBack and DLT 8000 for large file ( Oracle DB ) that gives a rate of 26 GB per Hour.

Magdi