Open karlmsmith opened 7 years ago
Comment by joe.mclean on 1 Nov 2004 23:47 UTC More info: Yonghua has encountered this bug. The following was not known before and may shed some light.
"this only happens when the [bad url] is the first dataset opened. If the [bad url] is on the second dataset opened, an error message can be displayed correctly."
DODS bugzilla http://dodsdev.gso.uri.edu/bugzilla/show_bug.cgi?id=716 is RESOLVED/WORKSFORME because James could not reproduce the problem.
Comment by @AnsleyManke on 2 Nov 2004 17:27 UTC Further information from Yonghua,
It looks to me that the "use" command first asks for the DAS of the specified dataset. The correct behavior for Ferret is to check if the returned DAS is correct, if the format of the DAS is wrong it will say:
** netCDF error: Attribute not found is this a CDF file ?
But somehow Ferret does not check the format of DAS if the "bad url" is the first dataset opened.
Comment by kevin-obrien on 2 Nov 2004 17:51 UTC I think this is a DODS problem. I get the following error when attempting to access the file using the latest DODS dncdump (with caching turned on) which I built from source using the most up-to-date code checked in on the DODS trunk:
./dncdump http://www.cdc.noaa.gov/cgi-bin/nph-nc/Datasets/ncep.reanalysis.derived/surface_gauss/skt.mon.inter.std.nc Caught error while initializing remote source: An internal error was encountered in HTTPCache.cc at line 1630: Could not open named cache entry meta data file. Please report this to support
@unidata
.ucar.edu
./dncdump: http://www.cdc.noaa.gov/cgi-bin/nph-nc/Datasets/ncep.reanalysis.derived/surface_gauss/skt.mon.inter.std.nc: Unknown error 1002
With caching turned off, the dncdump command works fine (as does the data access in Ferret). I will report this to James...
Comment by kevin-obrien on 3 Nov 2004 18:40 UTC In trying to simplify the problem, I have discovered the following:
Roland also had a bad URL: http://www.cdc.noaa.gov/cgi-bin/nph-nc/Datasets/noaa.ersst/sst.mnmean.nc
which is again at CDC. CDC is using dap server version 3.2.9. We at PMEL are using version 3.4.5. This may or may not matter....
It seems to be related to the .index file in the .dods_cache directory. My take on what is happening is that if the requested cdf file is referenced in the .index file, but the directories in the .dods_cache directory do not in fact have caching info for that file (or if perhaps it is out of date), the request fails. However, remove the .index file, and the request works fine. This indeed also happens if I do the following for our server:
rm -rf ~/.dods_cache ./dncdump -h http://www.ferret.noaa.gov/cgi-bin/nph-dods/data/coads_climatology.nc * works ** rm -rf ~/.dods_cache/* ./dncdump -h http://www.ferret.noaa.gov/cgi-bin/nph-dods/data/coads_climatology.nc Caught error while initializing remote source: An internal error was encountered in HTTPCache.cc at line 1630: Could not open named cache entry meta data file. Please report this to support
@unidata
.ucar.edu
./dncdump: http://www.ferret.noaa.gov/cgi-bin/nph-dods/data/coads_climatology.nc: Unknown error 1002
rm ~/.dods_cache/.index ./dncdump -h http://www.ferret.noaa.gov/cgi-bin/nph-dods/data/coads_climatology.nc * works again ***
This should be enough info to send to unidata support. I will send that today.
Reported by @AnsleyManke on 28 Apr 2004 17:56 UTC ! Found by Joe Mclean.
!The following DODS data set no longer appears in the DODS server directory.
! The problem is that Ferret v5.60 freezes up when trying to access it. This ! happens whether dods caching is on or off. It happens on both stout and ! jaguar. ctrl-C does not have an effect.
!dncdump v3.4 returns an error right away with caching turned on and off, both stout and jaguar. ! !*** DODS_Test/NETCDF/solaris/DODS_3.4/dncdump: ncopen failed on http://www.cdc.noaa.gov/cgi-bin/nph-nc/Datasets/ncep.reanalysis.derived/surface_gauss/skt.mon.inter.std.nc ! ! ! NOAA/PMEL TMAP ! FERRET v5.60 Linux(g77) 2.4.20 - 04/08/04 ! 26-Apr-04 11:23
yes? use "http://www.cdc.noaa.gov/cgi-bin/nph-nc/Datasets/ncep.reanalysis.derived/surface_gauss/skt.mon.inter.std.nc"
In totalview on jaguar, in cd_open_dset at the line cdfid = NCOPN(ds_des_name(dset), NCNOWRIT, cdfstat)
Totalview pops up a window saying, "you are about to throw and exception and teh process may run away. Do you wish to stop the process?"
Migrated-From: http://dunkel.pmel.noaa.gov/trac/ferret/ticket/865