Hi Mike,
Godin, Michael wrote:
Ethan,
Thank you very much. The example illustrates the use of collectionType within datasetFmrc, fmrcDefinition within aggregation, and the fmrcInventory element, which are otherwise unmentioned in the documentation.
It appears my problem was not using collectionType="ForecastModelRuns" within
datasetFmrc, since I am not using fmrcDefinition or fmrcInventory, and my catalog is now
working as one would expect, for DODS at least. I'm very excited to have this working.
Glad to here things are working.
Now onto WCS and NetcdfSubset services. WCS seems ok so far (I got a GetCapabilities
document), but NetcdfSubset was complaining that it can't open a file at
"/data/tmp/thredds/ncSubsetCache/<a very long name>.nc". Once I created a
tomcat-owned folder at /data/tmp/thredds/ncSubsetCache/ and took illegal filename characters
out of the datasetFmrc element's name, NetcdfSubset started to work -- hooray!
You can also change the cache locations in the threddsConfig.xml file
(in your content/thredds directory). There is a <dir> element in both
the <NetcdfSubsetService> and <WCS> elements. E.g.,
<NetcdfSubsetService>
<allow>true</allow>
<dir>/data/tmp/thredds/ncSubsetCache/</dir>
...
</NetcdfSubsetService>
<WCS>
<allow>true</allow>
<dir>/data/tmp/thredds/wcsCache/</dir>
<scour>15 min</scour>
<maxAge>30 min</maxAge>
</WCS>
More on the NCSS in the TDS reference
(http://www.unidata.ucar.edu/projects/THREDDS/tech/reference/NetcdfSubsetService.html)
and on the threddsConfig.xml at
(http://www.unidata.ucar.edu/projects/THREDDS/tech/reference/ThreddsConfigXMLFile.html#NetcdfSubsetService).
Constant Forecast Date is returning variable forecast dates, and it appears that having a
ragged time coordinate in the files messes with "Best Time Series" -- but these
are honest to goodness bugs -- it's really nice being at a point where I can experience
the bugs.
Can you explain what your ragged time coordinate looks like? We have it
working with data for models where different runs have different numbers
of forecasts. Is that where you are seeing the problem? It might also
depend on how the runs/forecasts are broken across files. Our system has
all forecasts from one run in a single file. But we have also worked
with data where each forecast is in a separate file.
Let us know what your setup is like where you see the problem (or, even
better, send us example files) and we can dig into it.
Regards,
Mike
Thanks,
Ethan
--
Ethan R. Davis Telephone: (303) 497-8155
Software Engineer Fax: (303) 497-8690
UCAR Unidata Program Center E-mail: edavis@xxxxxxxx
P.O. Box 3000
Boulder, CO 80307-3000 http://www.unidata.ucar.edu/
---------------------------------------------------------------------------
