Re: [fluka-discuss]: problem with FLUKA running

From: babintsev <babintsev_at_ihep.ru>
Date: Thu, 17 Sep 2020 07:42:30 +0300

Dear Paola,

  thank you for your instruction, but I have to use PC-cluster
  because of the very large amount of CPU time.

  The information for the last case with hanging job
   is placed at debug-job.txt  file :
  I asked system manager to debug my job...

  I send also my programs :
     source-test3.f
     mgdraw-test3.f
  and               test3.inp

  I have to give some commnets about programs :
  - I use 2 step methods : at the 1-st step I selected sample of
    particles
  - now I work with this sample .. BUT it is too small for my
    analysis to get good signal at the large distance (at my detectors)
          ct3, s1, s2,...
  - Using input sample of particles I generated another file
    where each particles was generated in small angle region around the
input
    vector ( small angle deviation with rndm method)
    So, my final file is now very large : 14 Gb !
    It was done in seperate program
    ... now I see that is was WRONG solution - I have to generate
    new particles inside the source.f ! ... but it is the next modification

  - so, from my source.f program I read ev by ev during job working
        for this I introduced  additional file : cycle.txt
    where I write the cycle number from which job must start.
    after each  complited cycle I write new number into this cycle.txt
    file.  So the next cycle FLUKA start from reading this cycle.txt
    and I know from which line I have to start to read input file
    for next stage of work ...

             ----------------
  Now my questions:

  - the job is hanging at thee strange instruction as it follows from
    debug-job.txt... it seems the system problems as you mentioned above.. ?
    ... how overcome them ?

  - I need to write another method to generated particles inside the
    source.f --> is it possible the change number of events per cycle
    inside the program ?
    ... may be there are some examples how to enlarge number of input
    particles ?

  - my next step will be : calculate efficiency of detector to register
    neutrons and other particles:
        Li6 glass scintillator  GS20
    may you know an example (idea) how to do it in FLUKA ?

              ------------------

  thank you
  Vladimir



13.09.2020 16:47, Paola Sala пишет:
> Hello
> It is difficult to help you without having the full information, in
> particular the source.f routine, and the exact answer from your system
> manager. Random hanging could come from interaction with the file system,
> or more generally with the operating system.
> If the code is still "running" you can try with gdb
> First you have to find the process ID (a number) of your running code: if
> you run it interactively (not on a batch queue) you can find it, for
> example, with (suppose your custom fluka executable is called myexe
> ps ax | grep myexe
>
> then, you can ask
> gdb myexe mynumber
>
> and you can discover which instruction is being processed
>
> Regards
> Paola
>>  Dear experts,
>>   I have problems in running jobs for FLUKA,
>>   i use 2-step method, at the 2-nd step I use  source.f
>>   to run secondary particles (large size !).
>>
>>   The problem : a lot of times FLUKA is stopped
>>         without any comments : it is not finished as usual
>>     BUT it is at the hanging state : ...it is waiting something....
>>
>>   At the 1-st time I asked system manager of our cluster  to look at
>> the job
>>   the answer was :
>>       it is waiting the answer from  "fuip"   .. or at "fuip" (may be
>> fiup ....)
>>              ... or something like this
>>       I do not remember for sure...
>>
>>   After that I re-run my job and it was finished.
>>   But now I using the same large input file and I cannot finish
>>   my job : every time FLUKA  get in the "hanging" state
>>              at DIFFERENT events (at different LARGE CPU time)
>>
>>   So, what I should do in this situation ?
>>
>>                 ------------------
>>
>>  Additional information :
>>
>>   cuts for my materials : see my input cards : test3.inp
>>    and files :  test3002.err  test3002.out
>>
>>                  ------------------
>>
>>  thank you
>>  Vladimir
>>
>>
>>
>
> Paola Sala
> INFN Milano
> tel. Milano +39-0250317374
> tel. CERN +41-227679148





__________________________________________________________________________
You can manage unsubscription from this mailing list at https://www.fluka.org/fluka.php?id=acc_info

Received on Thu Sep 17 2020 - 08:44:02 CEST

This archive was generated by hypermail 2.3.0 : Thu Sep 17 2020 - 08:44:05 CEST