[kepler-dev] [Bug 4254] New: - wrapping up, never ending workflow! (PN inside PN)

[bugzilla-daemon at ecoinformatics.org]

http://bugzilla.ecoinformatics.org/show_bug.cgi?id=4254

           Summary: wrapping up, never ending workflow! (PN inside PN)
           Product: Kepler
           Version: 1.x dev
          Platform: Other
        OS/Version: All
            Status: NEW
          Severity: minor
          Priority: P4
         Component: director
        AssignedTo: berkley at nceas.ucsb.edu
        ReportedBy: cxh at eecs.berkeley.edu
         QAContact: kepler-dev at kepler-project.org


[I'm going through my email logs looking for bugs that have
not really been solved.]

The problem is that a trivial example of PN inside PN
results in a "wrapping up" message and never terminates.

In January, 2009, Edward responded to email from Josep Maria Campanera.

Edward wrote:

> In theory, PN inside PN should behave exactly as if the inside PN director
> were not present. Why not just delete it?
>
> It is arguably a bug that the PN director doesn't tolerate
> PN opaque actors.  I think I tried to fix that once, but never
> got it fixed...
> 
> Edward


Josep Maria Campanera Alsina wrote:
> Dear all,
> I've found out that the error comes from the fact of using a PN
> director as an opaque director inside a worflow managed by another PN
> director! Somehow this problem is related to "Director compatibility"
> issue.
>
> So I've decided to simplify the workflow and use a model workflow just
> in order someone can find a trick to run it succesfully. The main idea
> is ¿How can we can continue the execution of a workflow after the
> execution of an opaque composite actors with PN director inside when
> the overall director is an PN one?
>
> In the attached example (exampleCompositeActors.xml), I cannot get a
> "Execution finished" but the Kepler shows the "wrapping up" signal or
> alternatively a permament "executing" signal although apparently
> nothing is running (only finished partly). No error message is
> displayed!
>
> Which worflow design I should adopt? What Am I doing wrong? How Can I
> pursue my aim to execute PN composite actors sequentially?
>
> All help appreciated!
>
> Josep Maria Campanera,
>
> PS1: I use Kepler Release 1.0.0 (May 2008) under Linux
> PS2: Thanks Edward for your comments,
>
> 2009/1/14 Edward A. Lee <eal at eecs.berkeley.edu>:
>> This is not a situation where it is refusing to terminate because of
>> unconsumed tokens. Typically, this is called by a deadlock. If you
>> are running on a Windows machine from the command line, you can get
>> a stack trace of all the executing threads by typing control-break.
>> If you forward that to this list, we can (maybe) identify
>> the culprit...
>>
>> Edward
>>
>>
>> Josep Maria Campanera Alsina wrote:
>>> Dear all,
>>> I get a "wrapping up" message from an execution of a workflow that
>>> coordinates the submission of computational chemistry jobs in a
>>> cluster queue system. The workflow is based on the
>>> TestJobFullCycle-SGE.xml provided by kepler itself, with some
>>> variations. Find attached the mentioned workflows.
>>>
>>> My workflow is a composite one. I'd like to execute sequentially
>>> several tasks: 1st) submit jobs to a cluster, 2nd) collect all outputs
>>> and save them locally and then 3rd) convert output files to other
>>> formats. Since some tasks are designed for PN director and some for
>>> SDF director the overall director is a PN director.
>>>
>>> However, the task under the PN director (in charge of job submission)
>>> never finishes tough the task is already completed and early shows the
>>> message "wrapping up", so the execution never finish step 1st and 2nd
>>> so the 3rd is never executed. Why PN director doesn't finish the task?
>>> How Can I stop the director and allow the workflow to continue?
>>>
>>> I think the problem is related to the PN director that as said in the
>>> user manual "may refuse to automatically terminate because tokens are
>>> always generated and available to downstream actors"! Here it comments
>>> that "One can correct the problem by changing the firingCountLimit
>>> parameter of the Constant actor to some finite value". If done that
>>> but it continues with the same problem.
>>>
>>> How can I overcome this? all comment will be appreciated!
>>>
>>> Thanks in advance,
>>>
>>> Josep Maria,