[IDV #LYT-733780]: .grb file on RAMADDA not working suddenly

Hi Jeff,

It looks like it's a 4.5.4 thing. I downloaded the 45GB file locally, and
4.5.4 sees something that triggers netCDF-Java to treat the file as a
multiple runtime file (maybe a bug...haven't dug deep). However, in
the 4.6 release, the file is handled as one would expect in which the
file is not read as a as two datasets (Best and 2D).

This would explain why Yuan was able to read it into the IDV without
any issues once the file was local (latest IDV uses netCDF-Java 4.6).

Sean

> Hi Jeff,
> 
> I just tried this with netcdfAll.jar v4.5.4 and v4.6.2, and it worked as
> expected on a sample grib1 file (it did not split the file into a TwoD and
> a Best collection).
> 
> I'll hop on my computer at work and start downloading the grib file and
> run my test code on it to see what happens.
> 
> In terms of GRIB, note that there are a ton of bug fixes in 4.6; you would
> likely see an increate in performance as well. Not sure how much work
> would be involved in bumping to 4.6, but that is what we recommend in
> general and is pretty much a necessity for grib files.
> 
> I'll et you know what I find out as soon as I am able to get the file.
> 
> Cheers,
> 
> Sean
> 
> > Hi Sean,
> >
> > I think we're using 4.5.4
> >
> >
> > Not sure what a partition type is. All I do is get the grids from the
> > GridDataset with:
> >
> > GridDataset.getGrids()
> >
> > Then to access the specific field:
> > GridDatatype grid = gds.findGridDatatype(var);
> >
> > -Jeff
> >
> >
> >
> > address@hidden> wrote:
> >
> > > Hi Jeff,
> > >
> > > What version of netCDF-Java are you using in this version of RAMADDA?
> > >
> > > It looks like the issue is that the grib file is being served up with the
> > > wrong partition
> > > type, which is exposing the dataset as both a 2D time dataset and a "best"
> > > dataset.
> > >
> > >
> > > http://www.unidata.ucar.edu/software/thredds/v4.6/tds/reference/collections/Partitions.html
> > >
> > > If you check out the .das, you can see what's going on there:
> > >
> > >
> > > http://weather.rsmas.miami.edu/repository/entry/show?entryid=synth%3A601aa404-ac59-4abb-8c4f-11c08474a214%3ALzIwMTEtMTEuZ3Ji&output=data.gridaspoint.form
> > >
> > > Notice every variable is repeated - one is for TwoD, and the other is for
> > > Best.
> > >
> > > What partition type are you specifying when the collection is created in
> > > the RAMADDA
> > > code?
> > >
> > > Sean
> > >
> > > > > Unidata folks - can you look into Brian's problem. You can see the
> > > subset
> > > > > form here -
> > > > >
> > > http://weather.rsmas.miami.edu/repository/entry/show?entryid=synth%3A601aa404-ac59-4abb-8c4f-11c08474a214%3ALzIwMTEtMTEuZ3Ji&output=data.gridaspoint.form
> > > > >
> > > > > Select one of the fields. The failure is showing up calling 
> > > > > GridDataset
> > > > > where it does not ind the named grid, e.g:
> > > > > GridDatatype grid = GridDataset.findGridDatatype("
> > > > > Geopotential_surface");
> > > > >
> > > > > -Jeff
> > > > >
> > > > >
> > > > Jeff and Brian,
> > > > I will move the ticket to the netCDF group for further help.
> > > >
> > > >
> > > > Yuan
> > >
> > > Ticket Details
> > > ===================
> > > Ticket ID: LYT-733780
> > > Department: Support netCDF Java
> > > Priority: Critical
> > > Status: Open
> > >
> > >
> >
> >
> > Hi Sean,
> >
> > I think we're using 4.5.4
> >
> >
> > Not sure what a partition type is. All I do is get the grids from the
> > GridDataset with:
> >
> > GridDataset.getGrids()
> >
> > Then to access the specific field:
> > GridDatatype grid = gds.findGridDatatype(var);
> >
> > -Jeff
> >
> >
> >
> > address@hidden> wrote:
> >
> > > Hi Jeff,
> > >
> > > What version of netCDF-Java are you using in this version of RAMADDA?
> > >
> > > It looks like the issue is that the grib file is being served up with the
> > > wrong partition
> > > type, which is exposing the dataset as both a 2D time dataset and a "best"
> > > dataset.
> > >
> > >
> > > http://www.unidata.ucar.edu/software/thredds/v4.6/tds/reference/collections/Partitions.html
> > >
> > > If you check out the .das, you can see what's going on there:
> > >
> > >
> > > http://weather.rsmas.miami.edu/repository/entry/show?entryid=synth%3A601aa404-ac59-4abb-8c4f-11c08474a214%3ALzIwMTEtMTEuZ3Ji&output=data.gridaspoint.form
> > >
> > > Notice every variable is repeated - one is for TwoD, and the other is for
> > > Best.
> > >
> > > What partition type are you specifying when the collection is created in
> > > the RAMADDA
> > > code?
> > >
> > > Sean
> > >
> > > > > Unidata folks - can you look into Brian's problem. You can see the
> > > subset
> > > > > form here -
> > > > >
> > > http://weather.rsmas.miami.edu/repository/entry/show?entryid=synth%3A601aa404-ac59-4abb-8c4f-11c08474a214%3ALzIwMTEtMTEuZ3Ji&output=data.gridaspoint.form
> > > > >
> > > > > Select one of the fields. The failure is showing up calling 
> > > > > GridDataset
> > > > > where it does not ind the named grid, e.g:
> > > > > GridDatatype grid = GridDataset.findGridDatatype("
> > > > > Geopotential_surface");
> > > > >
> > > > > -Jeff
> > > > >
> > > > >
> > > > Jeff and Brian,
> > > > I will move the ticket to the netCDF group for further help.
> > > >
> > > >
> > > > Yuan
> > >
> > > Ticket Details
> > > ===================
> > > Ticket ID: LYT-733780
> > > Department: Support netCDF Java
> > > Priority: Critical
> > > Status: Open
> > >
> > >
> >
> >

Ticket Details
===================
Ticket ID: LYT-733780
Department: Support netCDF Java
Priority: Critical
Status: Open