Job Submission Tutorial Grid Induction NA3 Training Team, Israel

After this tutorial you will be able to submit a simple job to the Grid and retrieve its output.

We will first learn how to submit jobs through the command line UI.
All commands and output will look like this

% This is the command This is the output

Prerequisites - a certificate, joining a VO and accessing a User Interface (UI) machine

There are 3 prerequisites for starting this workshop:

1. Receiving a certificate.
2. Signing for a VO.
3. Obtaining access to a User Interface (UI) machine.

Connecting to the UI

Users of Unix/Linux can login directly via ssh to the specific machine.

Windows users can use PuTTY or any other ssh terminal window such as Tera Term.

Before we begin, a clarification regarding passwords that are used throughout the tutorial:

There are typically two passwords involved.

1. The user and password of your account (your UI -the entering point to the grid).
2. GRID pass phrase - The password that protects the private key of the certificate.
   It is not necessarily the one given when exporting the certificate from a browser.
   This is the password needed in the command line in order to create a proxy certificate.
   The certificate (a file containing your private key, protected by password) is located at a directory called .globus under your root directory.

3. MyProxy pass phrase - The passphrase which relates to myproxy.
   The user will give on when initializing MyProxy (see MyProxy section)

Installing the certificate

Using the VO

Now that we have the certificate installed, we need to produce a proxy for running jobs through the command line UI. This proxy is used to authorize actions on all Grid elements (SE, CE, RB....) and avoids the need for sending your key over the network.

1. The command that creates a proxy that the UI machine will attach to any job you submit is

   % voms-proxy-init --voms MyVO Enter GRID pass phrase: Your identity: <your certificate details> Creating temporary proxy .................................................................. Done Contacting <voms server> Done Creating proxy ....................................................... Done Your proxy is valid until <12 hours from request>

   
   It will create a proxy valid for 12 hours. Uses the password of the userkey.pem.

   ---

   Note: VOMS means VO Membership System which enables a person to be a part of more than one VO (which was not possible in the past)
   This is why it is contacting the VOMS (a database) to verify your VO in the output and why you need to specify the --voms flag.
   

   ---

2. You can check your proxy status using
   

   % voms-proxy-info --all subject : <your certificate details with the additional specification that it is a proxy> issuer : <your certificate details> identity : <your certificate details> type : proxy strength : 512 bits path : /tmp/x509up_u1860 timeleft : <12 hours from request> === VO MyVO extension information === VO :MyVO subject : <your certificate details with the addiotional specification that it is a proxy> issuer : <your certificate details> with specification that it is voms attribute : /see/Role=NULL/Capability=NULL timeleft : <12 hours from request>

   
   Note: The --all flag is used to show the voms extension
   

Running a job

For now, let's start with a simple jdl we will call hostname.jdl:

As you can see, this jdl describes a regular job running a shell script.

• The executable is the shell and the argument is the simple shell script which sleeps for 5 seconds and returns the host name. The shell script has also to be transferred in the InputSandbox.
  Putting the shell script name as the executable is also possible.
  Pay attention that you do not need to put the executable in the InputSandbox, because there is a shell executable already present on the running CEs.
  A proprietary executable has to be included in the InputSandbox.
• The error and output are inserted to hostname.err and hostname.out files respectively and are destined to be retrieved through the OutputSandbox variable.
• The requirements of type other.GlueCEPolicyMaxWallClockTime are important when specifying the > sign.
  It ensures that if we have a long job, we will not arrive at a CE that removes too long jobs and get aborted.
• We've put the Virtual Organisation under which the job will run. We also want it to retry in case of failure at most 3 times.

Please create the previous two files anywhere under your home directory, via:

1. Editing tool such as VI, pico, emacs etc. You can paste it from this page.
2. Windows users can edit on their local machine and copy using, for instance, WinSCP.

1. There are two ways now to submit a job - the fast and annoying or the slower and easy.
   Every job submission need to delegate the proxy certificate to the broker called WMProxy.
   This can be done on each submission automatically, making it slightly slower or can be done once explicitly before all submissions.
   For doing it explicitly, choose a string identifier with the following command

% glite-wms-job-delegate-proxy -d MyString Connecting to the service https://wms.phy.bg.ac.yu:7443/glite_wms_wmproxy_server ================== glite-wms-job-delegate-proxy Success ================== Your proxy has been successfully delegated to the WMProxy: https://wms.phy.bg.ac.yu:7443/glite_wms_wmproxy_server with the delegation identifier: MyString ==========================================================================

2. The command for job submission is glite-wms-job-submit. The minimum parameters are
   1. the type of delegation of the proxy.
      either -d DelegIDString if you used glite-wms-job-delegate-proxy or automatic delegetion using -a flag
   2. the jdl name to run
   Submit your first job with the command (we'll use automatic delegation, but you can use explicit)
   

   % glite-wms-job-submit -a hostname.jdl Connecting to the service https://wms.phy.bg.ac.yu:7443/glite_wms_wmproxy_server ====================== glite-wms-job-submit Success ====================== The job has been successfully submitted to the WMProxy Your job identifier is: https://wms.phy.bg.ac.yu:9000/qTijWt8ofnnKzGU5RtxWwA ==========================================================================

   
   you get in response a job identifier in the form of a URI (as in the example above).
   Tip: use the -o MyJobIdFile if you don't want to remember the job ids and to be able to query all of them at once.
3. Check your job status until you get the status of Done using
   

   % glite-wms-job-status MyJobID ************************************************************* BOOKKEEPING INFORMATION: Status info for the Job : https://glite-rb.ct.infn.it:9000/bBo1R79o4R00pSceyRMo6w Current Status: MyJobStatus Submitted: Thu Aug 24 10:19:36 2006 CEST *************************************************************

   And when it's done you'll get

   ************************************************************* BOOKKEEPING INFORMATION: Status info for the Job : https://glite-rb.ct.infn.it:9000/bBo1R79o4R00pSceyRMo6w Current Status: Done (Success) Exit code: 0 Status Reason: Job terminated successfully Destination: ce01.afroditi.hellasgrid.gr:2119/jobmanager-pbs-see Submitted: Thu Feb 14 22:05:37 2008 IST *************************************************************

   
   
4. Now we can retrieve the files listed in OutputSandbox. Be aware that the directory specified will be created.
   If a directory already exists, you will receive a warning that it will be overwritten and a prompt asking whether you approve.
   Pay attention that the --dir flag must come before the URI (even when taken from a file using -i job.id)
   

   % glite-wms-job-output --dir output MyJobId; Connecting to the service https://147.91.84.25:7443/glite_wms_wmproxy_server ================================================================================ JOB GET OUTPUT OUTCOME Output sandbox files for the job: https://wms.phy.bg.ac.yu:9000/S0NjHO5qKS72qrPtRBYBLA have been successfully retrieved and stored in the directory: /home/<your account>/output *********************************************************************************

   
   
5. You can verify that the status has changed to CLEARED.
   

   % glite-wms-job-status MyJoBID ************************************************************* BOOKKEEPING INFORMATION: Status info for the Job : https://wms.phy.bg.ac.yu:9000/S0NjHO5qKS72qrPtRBYBLA Current Status: Cleared Status Reason: user retrieved output sandbox Destination: ce01.afroditi.hellasgrid.gr:2119/jobmanager-pbs-see Submitted: Thu Feb 14 22:06:37 2008 IST *************************************************************

That's it, You've submitted a job yourself !

Let's recap on the preparations and running:

Tip: If you don't want to type the VO name every time (with --vo option), you can set the following environment variable to your vo:

For more complex jobs such as DAG, Collection and Parametric jobs see the Advanced topics part of the tutorial.

Q: Hold on a second - what happens if my job is longer than the proxy life time (12 hours)?

A: Whoops. You're job is aborted. But have no fear, MyProxy is here.

Lets learn about MyProxy