Migrate on the cluster

Migrate version 2.1.3 is on the cluster.

You can run Migrate from a command line. To use it type "qrsh" (this will log you onto a single cluster node) and then type "migrate-n". Now use Migrate as you normally would in interactive mode.

A drawback of using Migrate this way is that if you log out of your qrsh session (or your cluster session) your Migrate job will be aborted.

You can get around this by using the runjobs.migrate script with a text file (with UNIX line breaks) that contains all of the commands that you would type in if you were running the program interactively. An example file of the command file is given below.

Here are the steps:

  1. Logon to the cluster.

  2. Change into the directory that has the file containing your Migrate commands (i.e., the command file).

  3. Type: "runjobs.migrate nameofcommandfile"

  4. Wait a few moments and then type Control-D. You will not receive a command prompt before executing Ctrl-D.

  5. The output that normally goes to the screen when running in interactive mode will be saved into a file called "nohup.out". This file will have no linebreaks, so depending on the program that you use to view it you may need to wrap the text.

Example of command file that is specified in step 3:

D
1
b
Y
i
1
/home/dsvd3/migrate/GaldoveWolfinfile.txt
7
/home/dsvd3/migrate/GaldoveWolf
y
p
1
o
6
etc.

It is important note that the full path is given for all files (infiles and outfiles). It is also important to note that the commands are case sensitive. Use blank lines to represnt "return" keystrokes.

In your home directory, you will notice files with the names "QRLOGINe_XYZ" and "QRLOGINo_XYZ", where XYZ is the job number of a particular Migrate job. These are the standard out and error files produced by the qrsh session. You can delete these when your job is finished. They may be helpful for troubleshooting purposes.