[kepler-dev] [Bug 3991] New: - large workflow errors

[bugzilla-daemon at ecoinformatics.org]

http://bugzilla.ecoinformatics.org/show_bug.cgi?id=3991

           Summary: large workflow errors
           Product: Kepler
           Version: 1.x dev
          Platform: Other
        OS/Version: Windows XP
            Status: NEW
          Severity: normal
          Priority: P2
         Component: general
        AssignedTo: berkley at nceas.ucsb.edu
        ReportedBy: soong at nceas.ucsb.edu
         QAContact: kepler-dev at kepler-project.org


I have a large workflow (both.xml) that seems to be generating these errors: 

Unable to parse the MetaData: null

The selected Entity in momol cound't [sic] be found

Error encountered in: <property name="endpoint"
class="ptolemy.kernel.util.StringAttribute"
value="http://ecogrid.ecoinformatics.org/knb/services/QueryService">
null

The error is not always the same, and is sometimes one and sometimes another. 
It seems to occur more frequently under 1.x dev than under 1.0.0.  Running
Kepler after cleaning the cache or switching between 1.0.0 and trunk seems to
sometimes but not always trigger it.  The problem seems to be diminished on
linux but does seem to occur.  Running Kepler with -Dmax=1024m does not seem to
fix the problem in Windows, but seems to help in linux.

Separating the composite actors into individual workflows, rain.xml and
buffalo.xml, seems to isolate the errors to buffalo.xml, but seems to make them
less consistent and more intermittent.

Some of the buffalo data is publicly accessible and some is not, so the example
workflows will generate permission errors.  The workflows do point only to the
accessible tables and should run, but the errors are thrown, regardless.

In addition, while investigating, I have been inconsistently getting startup
errors under Windows.  I suspect they might be due to my starting Kepler too
soon after closing it, after it returns to the command prompt but before it
truly closes.  One error is about failing to parse -kepler.  I seem to recall
this may have something to do with Kepler not being able to open some database
or other, but there are no other instances of Kepler visibly open this time.

I have also seen an error on startup about the user library not being
available.