Re: [thredds] NcML aggregation strangely pulling files from multiple datasets

• Subject: Re: [thredds] NcML aggregation strangely pulling files from multiple datasets
• From: John Caron <caron@xxxxxxxxxxxxxxxx>
• Date: Fri, 20 May 2011 08:37:12 -0600

On 5/19/2011 6:37 PM, John Maurer wrote:

Hi John:

On Thu, May 19, 2011 at 2:17 PM, John Caron <caron@xxxxxxxxxxxxxxxx <mailto:caron@xxxxxxxxxxxxxxxx>> wrote:

    Hi John:


    On 5/19/2011 3:24 PM, John Maurer wrote:

    Hi all,
    I'm seeing some weird behavior today with one of our joinExisting
    NcML aggregations for high-frequency radio (HFR). The OPeNDAP
    page lists variables that are from a totally different dataset
    (e.g. HFR does not measure water temperature):
    http://oos.soest.hawaii.edu/thredds/idd/hfr.html?dataset=kokagg.
    I looked at "content/thredds/cache/agg/" at the aggregation cache
    file for the dataset (hioos-hfr-kokagg) and it confirms that it's
    strangely pulling files from 3 unrelated datasets: a glider, an
    NCOM global FMRC, and the HFR data. Since nothing has changed
    recently in my config file for the dataset and I haven't recently
    upgraded TDS, this has me worried about a possible disk problem
    from recent power outages? The three datasets are not
    subdirectories of each other, so I don't see how TDS can be
    lumping them together. Any ideas what could be causing this?

    A brief look at what you put on ftp confirms it : its a mystery ;^(

    No chance you are playing games with symlinks is there?


Nope, no use of symlinks involved.

    In addition to the above, 5 of my FMRC's are suddenly
    inaccessible today as well and were recently working just fine.
    It would seem their file aggregations may have similarly gone
    haywire as well, but their cache files under
    "content/thredds/collection/" (right?) are binary (.jdb) so I

can't confirm whether they're combining multiple datasets.

    If you have remote debugging on (and you should!) you can look at
    the files with:

    https://server/thredds/admin/debug?Collections/showFmrcCache
    
<https://motherlode.ucar.edu:8443/thredds/admin/debug?Collections/showFmrcCache>

Yep, I later tried that as well. For all of the FMRC's that were failing, the cache had not been updated for several days.

    The catalogs give an error in threddsServlet.log with the message:
            at
    ucar.nc2.ft.fmrc.FmrcDataset.getNetcdfDataset2D(FmrcDataset.java:166)
            at ucar.nc2.ft.fmrc.Fmrc.getDataset2D(Fmrc.java:183)

    can you send the file content/thredds/logs/featureCollectionScan.log ?

Sorry, I would send that log except that it seems it was overwritten when I restarted TDS so I can't find the same time period and error messages anymore. Here are the relevant snippets that I dug out earlier during my investigation. TDS can't seem to find files anymore, although the directories and filenames are all correct and haven't changed for quite some time:

[2011-05-19T12:21:46.145-1000] ERROR ucar.nc2.ft.fmrc.FmrcDataset: Fmrc collection is empty =fmrc:NCOM Hawaii Regional Ocean Model [2011-05-19T12:21:46.145-1000] ERROR ucar.nc2.ft.fmrc.Fmrc: /export/lawelawe1/model/ncom/

hawaii/**/ncom_relo_hawaii_u_#yyyyMMddHH#_t.*\.nc$: makeFmrcInv failed

did you send the catalog that has this fmrc in it? if not, can you send?

• References:
  • [thredds] NcML aggregation strangely pulling files from multiple datasets
    • From: John Maurer
  • Re: [thredds] NcML aggregation strangely pulling files from multiple datasets
    • From: John Caron
  • Re: [thredds] NcML aggregation strangely pulling files from multiple datasets
    • From: John Maurer