The main source of information about Tramontana is provided by the PISA WIKI located at the web pages http://wiki.infn.it/cn/csn4/calcolo/csn4cluster/home while more dettailled information on cluster are HERE. Fast access to the cluster status can be visualized:

• USAGE
• LSF Monitoring

The Tramontana cluster can be accessed also using a simplified access to the grid provided by the web-portal created by CNAF IGI grid portal

The first step is to be authenticate to its usage

voms-proxy-init -voms theophys:/theophys/IS_OG51/Role=parallel  --valid 24:00
myproxy-init -d -n  -c 3880

An easy access to the file on the cluster can be achieved using the "uberftp" protocol

alias UBER='uberftp -D 0 gridce3.pi.infn.it '
UBER

As "theophys" user we do have to three-storage areas whose actual usage can be tested looking at the following files

uberftp -D 0 gridce4.pi.infn.it "cat /gpfs/ddn/csn4home/.disk_usage"
uberftp -D 0 gridce4.pi.infn.it "cat /gpfs/ddn/srm/theophys/.disk_usage"
uberftp -D 0 gridce4.pi.infn.it "cat /gpfs/ddn/theompi/chk/.disk_usage"

• The first storage area is for creating local job tree and usually is content is erased after job completion
• The second storage area is for permanent STORM storage and its content can also accessed using POSIX standard
• The third area is a scratch area were file may live for ever and can be used to save JOB checkpoint or partials job results. We save our works on the sub-tree "/gpfs/ddn/theompi/chk/OG51/Parma" and in its subdirectory "BarModeRuns" all the temporary file produced by the run project "BarMode". Similar convention will be used for any other project.

For project involving Cactus runs we decided that (let us suppose that the main directory of the project is: "/gpfs/ddn/theompi/chk/OG51/Parma/BarModeRuns") that this tree will contain a

• ./par a par directory were the parameter file used to run the simulations are stored.
• ./[simname] A directory containing the results obtained running the par file ./par/[simname].par
• ./[simname]/CHECKPOINT The checkpoint of the various restarts
• ./[simname]/output-0000 The output of the first start of the runs
• ./[simname]/output-0001 The output of the second start of the runs (from checkpoints)
• ./[simname]/output-….
• ./[simname]/output-N The output of the Nth-start of the runs

We have prepared a … THE BEST WAY to manage jobs is using the following commands (SUBMIT/STATUS/GET results)

 glite-wms-job-submit -a -o A100M15b265_r50_8.job A100M15b265_r50_8.jdl                   [SUBMIT]
 glite-wms-job-status -i A100M15b265_r50_8.job                                            [STATUS]
 glite-wms-job-cancel -i A100M15b265_r50_8.job                                           [CANCEL]
 glite-wms-job-output --noint  -i A100M15b265_r50_8.job --dir  ./A100M15b265_r50_8.dir    [GET OUTPUT]

That's all folks….

Now the compilation on "Tramontana" is in two steps. The first step is to create a tar ball of the compilation tree and transfer it on the PISA SRM (Storage Resource Manager). The second step is to submit a compilation job on cluster. This stage will be revised in near future to do compilation on the pisa user interface.

To check which exe is on the SRM one can gives the commands:

lcg-ls -l srm://gridsrm.pi.infn.it/theophys/IS_OG51/Parma/Cactus/exeTRAMONTANA/
lcg-ls -l lfn:/grid/theophys/IS_OG51/Parma/Cactus/exeTRAMONTANA 

lcg-ls -l  srm://gridsrm.pi.infn.it/theophys/IS_OG51/Parma/SRC/
lcg-ls -l  lfn:/grid/theophys/IS_OG51/Parma/SRC/

The one with name ending with "_MP" will refer to executable with OpenMP.

All the script and data need to make compilations are stored in the directory: "/storageQ01/BarModeProject/TramontanaSubmitions/Compilation" on einstein.pr.infn.it.