[kepler-dev] adding new job manager

Norbert Podhorszki pnorbert at ornl.gov
Wed Apr 30 07:03:40 PDT 2008


Hi Ufuk,

This happens if your umask prohibits –x for the files. The scp creates the
file on the remote host with permissions according to umask.
You can try:
1. set your umask to –x friendly (on your ‘remote’ machine). Mine is 0022.
2. or use “local” instead of “karetta.be.itu.edu.tr” in the JobManager’s
target parameter (this will let the whole job class work with local java
operations instead of ssh)
3. or put your executable on the remote machine and enumerate it in the
‘remotefiles’ input for the JobCreate actor instead of the ‘executable’
input (the ssh operation will perform a cp command on the remote machine,
which preserves permissions)

Of course, it’s a bug. Thanks for pointing it out.

Norbert

  Norbert Podhorszki
  ------------------------------------------
  Scientific Computing Group
  National Center for Computational Sciences
  Oak Ridge National Laboratory
  Phone: (865) 574-7159




On 4/30/08 8:40 AM, "Ufuk Utku Turunçoğlu" <turuncu at be.itu.edu.tr> wrote:

> Hi,
> 
> Thanks for your help.
> 
> By the way, before adding LSF actor, i also install CONDOR into my laptop to
> working on my workflow. So, I think it could be a problem in coping executable
> file into the $(user_name)_$(date) folder (in my case it is
> turuncu_Apr30_151053EEST_2845 and s generated automatically by KEPLER) in job
> submission process. When i run the simple job submission workflow, executable
> which is defined in JobCreator actor is copied into automatically generated
> folder without executable permission and CONDOR job gives,
> 
> Hold reason: Error from starter on slot1 at karetta.be.itu.edu.tr: Failed to
> execute 
> '/Users/turuncu/Documents/workspace/wf-phd/condor/turuncu_Apr30_151053EEST_284
> 5/simple.x' with arguments 4 10: Permission denied
> 
> error. The orginal file permission is,
> 
> -rwxr-xr-x   1 turuncu  staff  20224 Apr 30 11:15 simple.x
> 
> and the copied one is,
> 
> -rw-r--r--  1 turuncu  staff  20224 Apr 30 15:10 simple.x
> 
> My script and executable files is in
> /Users/turuncu/Documents/workspace/wf-phd/condor and the definition of the
> parameters in JobCreator actor is,
> 
> cmdfile = "/Users/turuncu/Documents/workspace/wf-phd/condor/simple.submit"
> executable="/Users/turuncu/Documents/workspace/wf-phd/condor/simple.x"
> workdir="/Users/turuncu/Documents/workspace/wf-phd/condor"
> 
> PS: My simple job submission workflow also contains JobManager and
> JobSubmitter.
> 
> Best wishes,
> 
> --ufuk
> 
> On Apr 30, 2008, at 3:06 PM, Norbert Podhorszki wrote:
> 
>>  Hi Ufuk,
>>  
>>  The JobManager.java: preloadSupporterClasses() loads the JobSupportXXXX
>> classes, so add your class there.
>>  In the workflows (JobManager actor), you have to use the XXXX for naming
>> your support class.
>>  Finally, workflows/test/job contains workflow examples for job controls.
>>  
>>  Best regards
>>  Norbert
>>  
>>    Norbert Podhorszki
>>    ------------------------------------------
>>    Scientific Computing Group
>>    National Center for Computational Sciences
>>    Oak Ridge National Laboratory
>>    Phone: (865) 574-7159
>>  
>>  
>>  
>>  On 4/29/08 8:01 PM, "Ufuk Utku Turunçoğlu" <turuncu at be.itu.edu.tr> wrote:
>>  
>>  
>>> Hi,
>>>  
>>>  I try to implement Platform LSF (Load Sharing Facility) support to KEPLER
>>> Job Manager actor. I found that types of job managers is controlled by
>>> src/org/kepler/job/JobSupportXXXX.java and
>>> src/org/kepler/job/JobManager.java files. What steps is necessary to adding
>>> new job manager? Is there any other file that control it.
>>>  
>>>  best regards,
>>>  
>>>  --ufuk
>>>  
>>>  
>>  
>>   
> 
> 

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mercury.nceas.ucsb.edu/kepler/pipermail/kepler-dev/attachments/20080430/9c782133/attachment.html>


More information about the Kepler-dev mailing list