Using Data Carousel on CCJ-HPSS

This document is draft version, the content will be changed without notice in near future.

User setup (easy way to retrieve data from the HPSS)

before starting
First of all, users login to any of the linux node, or ccjsun. The carousel scripts can be executed from any of these machines. Before running the scripts, the users must issue the command,

% ssh hpsss0.riken.go.jp

Note that you do not need to actually login hpsss0 (You probably do not have an account on this machine). This command will add the proper host key for hpsss0 to your ~/.ssh/known_hosts file.

File List

Next, users must creat a file which contains the list of files. The format of this list would be like,

[filename(with full path) in HPSS] [pftp://username@machine/output_filename]
.....(same format)....
.....(same format)....
.....(same format)....
Note: the character jusr after machine name is not ": colon" but "/ slash" against your usual expectation.
for instance, we will name the list file, "input1.lis", and it contains the request,

/home/phnxsink/Run-01/DST/v03/DST_v03_Stream01-0000007688-0000.proot pftp://jinnai3@ccjsun.riken.go.jp/ccj/w/r01/jinnai3/dc_test/DST_v03_Stream01-0000007688-0000.proot
next example (named as "input2.lis") contains the 5 files to be transferred:

/home/phnxsink/Run-01/DST/v03/DST_v03_Stream01-0000007688-0000.proot pftp://jinnai3@ccjsun.riken.go.jp/ccj/w/r01/jinnai3/dc_test/DST_v03_Stream01-0000007688-0000.proot
/home/phnxsink/Run-01/DST/v03/DST_v03_Stream01-0000007689-0000.proot pftp://jinnai3@ccjsun.riken.go.jp/ccj/w/r01/jinnai3/dc_test/DST_v03_Stream01-0000007689-0000.proot
/home/phnxsink/Run-01/DST/v03/DST_v03_Stream01-0000007692-0000.proot pftp://jinnai3@ccjsun.riken.go.jp/ccj/w/r01/jinnai3/dc_test/DST_v03_Stream01-0000007692-0000.proot
/home/phnxsink/Run-01/DST/v03/DST_v03_Stream01-0000007693-0000.proot pftp://jinnai3@ccjsun.riken.go.jp/ccj/w/r01/jinnai3/dc_test/DST_v03_Stream01-0000007693-0000.proot
/home/phnxsink/Run-01/DST/v03/DST_v03_Stream01-0000007694-0000.proot pftp://jinnai3@ccjsun.riken.go.jp/ccj/w/r01/jinnai3/dc_test/DST_v03_Stream01-0000007694-0000.proot
The location of these files (the directory structures and absolute paths) can be checked here, and respective file names can be looked up to with cftp command.

In above cases, the destination machine is declared as ccjsun.riken.go.jp, in accordance with the logined machine. (From linux nodes, you can indicate the destination to their local disk, i.e. /job_tmp , see the description in 4. User Accounts of ccj user guide for /job_tmp usage. You should use the local disk of your logined machine to avoid writing over NFS.)
We can use the abbreviated form of file paths(don't need to specify long absolute path name for all files.), detailed description can be found in this document.

Running the scripts,

Next, you can excute the carousel script as,

% /usr/local/carousel/hpss_user.pl -f input1.lis

If you encounter the error message, please let jinnai3@riken.go.jp know
the situation.

this will submit the request in MySQL server (in current configulation, ccjsun is working as Mysql server.), and create the entries in database phenix_carousel .

On the server end(hpsss0.riken.go.jp), the common account ccjrdat runs cron job, which invokes the server script hpss_server.pl for every 12 minutes.
This script inspects the Mysql database on ccjsun, and generates the new list of input files (this list includes the file names from all-users in order to prevent chaos in data retrieving.). This new list is submitted to ORNL batch system on HPSS server machine.

Therefore if it goes well, you can retreive your requested files within about 15 minites!


Adress Rad-Lab, RIKEN,2-1, Hirosawa,Wako, Saitama,351-0198, JAPAN
TEL +81-48-462-1111 ext.3279
FAX +81-48-462-4641 (office RIKEN-Wako.)
e-mail jinnai3@riken.go.jp