We will meet at 8am pacific time. I have booked the VRVS (not EVO) Geminga room.

Status

  • The job submission, pipeline and database has been modified to accept non-numeric job ids.
  • Currently the pipeline needs to be statically configured to submit either to SLAC or IN2P3
    • We have a test version of the pipeline running configured to submit to IN2P3
    • We are working to make it possible for one instance to submit to multiple sites
  • The job submission daemon is running at IN2P3.
    • There are some remaining issues to be solved
  • NFS space is available at IN2P3 for storing log files, temporary data etc.
    • GPLtools and GR-v10r1 are installed at IN2P3

Outstanding Technical issues

The command is not correctly being sent to the batch job

For instance currently we need to prepend -s to the command when sending it to BQS.

The job name is not correctly set yet.

Current the BQSJobControlService.submit() returns the "name" of the job (as set by Job.setName()). This is probably incorrect, it should return a unique jobID for the job (e.g. 10496.ccwalm14)? (I am assuming this is what needs to be passed to BQSJobControlService.status() and BQSJobControlService.cancel(), is this correct?)

Mystery of why we need to include BQSJobControlStub into the jar file

This may be related to the way the server is being started. In my case I start it with a script like:

#!/bin/csh
setenv CLASSPATH org-glast-jobcontrol-1.0.jar
# Make sure files created by batch jobs are group writable
umask 2
rmiregistry &
mkdir -p /u1/tmp/bsub
java -Djava.util.logging.config.file=logging-new.properties org.glast.jobcontrol.lsf.JobControlService

It would be interesting to know how it is stated at IN2P3.

We need to set up rsync or equivalent so we can view log files and other output from IN2P3

Not sure what the status of this is?

Need more testing, including testing retrieving status of running jobs.

Mechanics of running pipeline tasks at IN2P3

  • Initially we should try reproducing some existing task at IN2P3
    • What would we be an appropriate task?
    • We need to add a step to copy data back from IN2P3 to SLAC
      • Perhaps using xrootd?
  • Who will be responsible for running jobs at IN2P3?
  • How will we get new versions of GlastRelease, GPLtools etc installed at IN2P3?

Tony's notes from meeting

  • How much space is available in NFS?
  • Is there any local scratch space available at IN2P3?
  • We need to find out if xrootd works at IN2P3?
    • There is a minimal file size for HPSS, 20MB.
  • Current jobs produce
    • Merit: 1.5MB
    • MC: 93MB
    • Current jobs simulate 4 orbit seconds, 90mins on most recent box.
  • Need general information about Lyon farm.
    • Claudia will send web site.
  • Will there be any specific quota for GLAST jobs
    • Maybe we should be proactive about what we need
  • Do we need a local account for Tom and others
    • Claudia will check if she is allowed to send us the password for the glastpro account at IN2P3.
    • Or find out if Tom already has an account
  • Background V11R2 would be a good test
    • Claudia will install V11R2
      • It would be good to try to use the binaries
        • Claudia will let us know if she has problems

Answers to some questions

How much space is available in NFS?

We were authorized to use up to 150 Gb for GCRCalib + PipelineII needs. I have now used about 3 Gb for PipelineII (intallation of different GR versions and GPL tools). For the moment, it remains about 55 Gb for PipelineII future needs.

Is there any local scratch space available at IN2P3?

Regarding SCRATCH space those three classes are comparable to the scratch space at SLAC (4-5 Gb).

Class

Scratch Space (Mb)

 

Default

Maximum

A

400

5

150

G

1025

10

250

T

2050

20

500

We need to find out if xrootd works at IN2P3?

HPSS storage system file size constraint:

Files beeing stored in HPSS system must be larger that 20Mb.

Need general information about Lyon farm.

See http://cctools.in2p3.fr/mrtguser/info_bqs_class_config.php

  • No labels