[cdelgadop]

Hi

Let's say i have a program that creates several files while running and i want to move those files to a remote server without actually store those files locally cause i don't have enough space. I'm thinking about tar pipes with rsync or netcat commands but i have tried these approach before with commands whose result were just one file (like tar or pax commands), something like this:

cd /sourcedir; tar -cvf - | nc -q 10 -l -l 7878

then in the remote server:

cd /targetdir; nc remoteserver 7878 | tar -xvf -

This somehow have worked for me. I have even used rsync and ssh tunneling in order to make such transfer as secure as possible.

Now i want to use the db2move command in DB2 (whose creates several files) in order to create all output files needed to migrate our DB2 Datawarehouse of about 1TB based on Linux on Mainframe to one LPAR on AIX. I don't have 1TB of aditional space in the Mainframe so i need to use the pipe feature of UNIX to move those N files resulting of the runnning of db2move to AIX through the network.

Thanks in advanced

[pedz]

Well, first. Your "working" example isn't really working is it? All the data is stored on both machines -- at least until you remove it from the source machine. Right?

With PostgreSQL, you can do what db2move is doing from a client machine. You simply specify the hosting machine, port, and db name. I bet db2 has a similar function. That implies that you have to have the db2 client software on the client machine. There may be a licensing issue there.

Another option is NFS. Create and export a directory on the client machine. This machine will be the NFS server. Then from the db server machine, mount that directory into the file system. Now, do the db2move command but have it go to the mounted NFS directory. The files, in this case, never touch the db server's file system but are sent to the NFS server's file system as they are created. You can use DFS, AFS, GPFS, etc instead of NFS.

I would go with using the db2move command from the client if at all possible. I bet it is.

[esanchezvela]

Originally Posted by cdelgadop

Hi
Let's say i have a program that creates several files while running and i want to move those files to a remote server without actually store those files locally cause i don't have enough space. I'm thinking about tar pipes with rsync or netcat commands but i have tried these approach before with commands whose result were just one file (like tar or pax commands), something like this:

Now i want to use the db2move command in DB2 (whose creates several files) in order to create all output files needed to migrate our DB2 Datawarehouse of about 1TB based on Linux on Mainframe to one LPAR on AIX. I don't have 1TB of aditional space in the Mainframe so i need to use the pipe feature of UNIX to move those N files resulting of the runnning of db2move to AIX through the network.

Thanks in advanced

Would it be fair to say that all you want to do is to create a db2dump and send it to db2 on AIX???

sometimes I do this...

ssh remuser@remhost mysqldump <rem_database> | mysql <local_db>, this allows me to replicate my remote database to the localhost where the hosts are connected by a limitting socks server which does not allow mysql replication.

would it be possible to create a dump of the database and send it to stdout?

A different solution would be to create a tunnel between Linux and AIX using ssh using remote port forwarding, establish a session from the linux box to AIX establishing port-forwarding on the command line, next dump the db2 database to stdout and establish a connection to the aix db2 instance using the local port.

linux> ssh -L 8080:127.0.0.1:<db2port_on_AIX> remuser@aixhost
On a different terminal session:
linux>db2dump database | db2 --port 8080 (you'll have to specify the login credentials here for the AIX instance of the database> remdb

Now, this and the netcat solutions should work fairly well and are kind of cool to implement, but they will not be supported by IBM and I would like to believe that if your customer is running Linux on the mainframe and the database is close to 1 TB on each side of the equation, they would expect a commercial level type support for their data replication.


DB2 must have replication tools specifically desgined for this purpose and supported.

best regards,
esv.

[cdelgadop]

Hi

Not in this case. This is not a dump of the DB2. If such case i'd just take that dump and send to TSM and restore in AIX. We are here changing from Mainframe to PowerPC, we are changing binaries. Pretty much the same as a SAP Heterogenous Copy from different HW. When you run a (some other DB) dump one file is created and you can move that file through a pipe, in this case db2move will create several files needed to create the DB in the AIX.

About DB2 replication tools, according to the IBM web site there's no way to migrate from different platforms unless using db2move command. Again, db2move creates several files, and i don't have enough space to hold them in the mainframe. I'm thinking about using a NFS FS.

Thanks for your answer.