Closed ocehugo closed 4 years ago
we are aware of the problem and need to regrid data on the proper grid. these files were not generated @ UWA but @ emii and have several issues, including non-up-to-date Conventions (CF-1.5,IMOS-1.2) and non consistent QC tests - flags
same would apply for COF data, while all data for ROT, SAG, TURQ have been upgraded recently
@scosoli , thanks for the quick reply.
I will update the title to include COF too. Let me know if you need anything from us.
thanks hugo - my only question would be, by when would you need the updated data available? will prioritise this task over others if needed
@scosoli,
I just checked and apparently the COF dataset can be aggregated - Hence, no grid problems. I'm removing it from the title to avoid confusion. Did you mean COF only need convention updates? If so, I think a new issue is the best thing to do.
my only question would be, by when would you need the updated data available? will prioritise this task over others if needed
it's not my call to define priority - @ggalibert!?
If I understand well this corresponds to 10 days of data that needs re-processing by the facility. It is quite small so maybe not necessarily a high priority but in the meantime it is breaking the aggregation service from the portal: any aggregation request for this site that includes [2008-06-02; 2008-06-12] will fail.
@scosoli , would that be OK to remove these files from THREDDS/S3 (we can still keep a copy somewhere) while we're waiting for you to provide us with re-processed datasets? You could prioritise this task to whenever would work best for you.
let me download these data first then you can remove access from thredds/S3. will notify you when done. We've found something funny on DM FV01 WERA radials after 201711070415 for all WERA sites, too. will open a separate ticket
There is a couple of issues here with the DM data set. 1, the same grid inconsistency is also found for RT data; 2, the grid seems to randomly change from dimension 80x80 to 64x72 for RT data 3, the FV01 DM data gridded on the 80x80 grid do not make sense. there is a clear mistake during the reshaping of a matrix. see attached godiva maps for the same data, the first for RT and the second for DM data
sorry, attached the same plot twice. the correct image for the 'wrong' DM FV01 is below:
the pattern shown in the second (third) plot is consistent within the entire time frame listed by @ocehugo above. @ggalibert I think you should remove with no hesitation the FV01 data while we fix them from scratch.
FYI: I found the same issue at other nodes (TURQ, for instance), but I had the chance to reprocess (and regrid) data consistently
Thanks @scosoli.
FV01 Affected files unindexed & removed.
PS: I assume the FV00 would be re-uploaded/overwritten together with the new FV01!? I didn't remove since they don't break anything but I'm willing to do so if helps to maintain consistency.
@ocehugo , yes, I will update also FV00 because they need to be regridded for consistency with other FV00 products.
@ocehugo, FV01 data for the subset listed above have ben regridded, QC'd and are ready to be uploaded. however, there may be an inconsistency before/after 2014q1 - not in terms of grid size but rather in terms of flipped dimensions. data need to be flipped upside/down (latitude), and left/right (longitude, GDOP, UCUR/VCUR, UCUR_sd/VCUR_sd, NOBS1/NOBS2, UCUR_quality_control/VCUR_quality_control. data can still be concatenated with ncrcat with no errors but aggregation would produce inconsistent results. It would be better also to run the full QC set to the entire record - has been done only for TURQ, ROT, SAG, BONC, not yet for CBG and COF. will do asap
@scosoli, thanks for the heads up.
I cannot locate the above files or logs of the upload. Did you really upload!?
IMO, you may wait to upload the entire record reprocessed. I think if the files are inconsistent with the rest it may cause more problems for the end-user than having no files.
PS: We don't use ncrcat for aggregation but java libraries. I think that if the files are inconsistent as you describe the aggregation will continue to fail or produce the same results you experienced with ncrcat.
hello @ocehugo , no I have not uploaded files yet. was waiting for some feedback. will reprocess all data consistently so to avoid possible issues with concatenation and also make sure all data are QC'd consistently.
@ocehugo, FV01 data for the time period (2007q4 to 2013q3), and also the files listed above, are ready for upload to the portal. will soon start uploading
@ocehugo @ggalibert , all data for the time period 2007Q4 to 2013Q4 are currently being uploaded (including the dataset listed here). grid issues should be fixed now and all data QC'd consistently with updated conventions and QC flags
@scosoli - I'm watching it and will verify tomorrow if we got any problems along the way.
thanks
update: I will check again on Monday - still about ~20hours to go given every file is taking about ~3sec.
@ocehugo I believe all files have been uploaded. I did some aggregation tests and everything seems to be working fine. if you confirm this too, I think we can close this issue
Yeap - all good. I tested earlier this week but forgot to reply here.
Thanks
Problem: We still got some files with a different grid for the ACORN CBG collection. This is blocking the aggregation service, since the grid on those files is different from others. Blame files are between 2008-06-02 and 2008-06-12 and were uploaded in Oct/2015.
Solution: Re-upload corrected files.
@scosoli, may you check/reupload the files below!?
Files below do not follow the standard Lon/lat grid dimensions:
List :