[IPT] [EXTERNAL] Re: How does one upload large datasets to GBIF?

Tue Jul 7 17:19:20 UTC 2020

Hi Annie,

With additional RAM allocated, the IPT can publish proportionally larger 
datasets.  However, this can be inefficient (or expensive in terms of 
RAM), especially if the dataset has extensions.

To construct the DWCA outside an IPT you will need:

- data files.  It's a good idea to check them for common errors -- 
incorrect number of columns, duplicate occurrenceIds and so on (the IPT 
does several checks like this).

- a meta.xml data description file, linking columns to Darwin Core 
terms. This can be written by hand, using various programming languages, 
or (often easiest if the process isn't to be repeated) by using an IPT 
to make a suitable mapping and extracting the resulting file.
- an eml.xml metadata file, describing the dataset.  The same applies 
here -- the IPT is useful for providing a UI to write this metadata, 
especially if all 8 are similar.

Once the DWCA exists, it should be copied to a webserver.

Note that using the registry API is not strictly necessary, and a 
publisher with a small, unchanging number of datasets outside the IPT 
need not use it.  They can simply give the helpdesk a URL for each 
dataset's DWCA file, and update the DWCA files at those URLs as necessary.

Using the API is useful for adding additional datasets, making changes 
(e.g. changing the URL) of the existing 8, or prompting GBIF to 
reprocess a dataset.  To use the API the technical team should create a 
suitable username (e.g. "usgs" or "bison") on both gbif.org and 
gbif-uat.org.  The latter is our test system.  They should then contact 
helpdesk at gbif.org to ask for permission for that account to make changes 
under the USGS 
<https://www.gbif.org/publisher/c3ad790a-d426-4ac1-8e32-da61f81f0117> 
publisher, or whichever publisher is/are appropriate.  This will only be 
on the test system at first.

It's then possible to register a new dataset under that publisher, 
following the example here: 
https://github.com/gbif/registry/tree/master/registry-examples/src/test/scripts 
and see the result.

For general questions on this, the GBIF API mailing list is probably 
most appropriate: https://lists.gbif.org/mailman/listinfo/api-users

If you have problems or errors with a specific dataset, 
helpdesk at gbif.org will be the best contact.  (They also read both 
mailing lists.)

Cheers

Matt

On 07/07/2020 17:48, Simpson, Annie wrote:
> Thank you, Laura, for your replies.
>
> The datasets have been exported from databases and cleaned. They are 
> generally UTF-8 tab delimited files. So it seems that the GBIF 
> Registry API would be the correct solution.
>
> We currently have 8 of these large datasets, only 2 of which would not 
> be updated in the future. Do you have names of GBIF Product Team 
> Members whom my technical team should contact to begin this process? 
> Is there "how to" documentation you can point me to that they should 
> read first?
>
> Annie
>
> ------------------------------------------------------------------------
> *From:* Laura Anne Russell <larussell at gbif.org>
> *Sent:* Tuesday, July 7, 2020 11:17 AM
> *To:* Simpson, Annie <asimpson at usgs.gov>; ipt at lists.gbif.org 
> <ipt at lists.gbif.org>
> *Subject:* [EXTERNAL] Re: [IPT] How does one upload large datasets to 
> GBIF?
>
> * This email has been received from outside of DOI - Use caution 
> before clicking on links, opening attachments, or responding. *
>
>
>
> I could also mention that it is possible to script the creation of the 
> Darwin Core Archives and then use the GBIF Registry API for the 
> connections with GBIF. Symbiota, PlutoF and some others are 
> successfully doing this. It does require some initial coordination 
> with our Product Team on how to set up and coordinate the registration 
> process and potentially with our Informatics Team.
>
> Best,
>
> Laura
>
> Laura Anne Russell
>
> Programme Officer for Participation and Engagement
>
> Global Biodiversity Information Facility (GBIF) Secretariat
>
> larussell at gbif.org (email)
>
> laura.anne.russell (Skype)
>
> @pagodarose (Twitter)
>
> #CiteTheDOI @GBIF
>
> https://www.gbif.org/
>
> +45 35 33 35 51 (office, direct line)
>
> GBIF
>
> Universitetsparken 15
>
> DK-2100 Copenhagen Ø
>
> Denmark
>
> *From: *IPT <ipt-bounces at lists.gbif.org> on behalf of "Simpson, Annie" 
> <asimpson at usgs.gov>
> *Date: *Tuesday, 7 July 2020 at 16.48
> *To: *"ipt at lists.gbif.org" <ipt at lists.gbif.org>
> *Subject: *[IPT] How does one upload large datasets to GBIF?
>
> Colleagues:
>
> What is the easiest or most popular way to send large datasets to 
> GBIF, ones that are too large for the IPT software (I think that is 
> more than 100MB zipped, 10+million records)? Does one modify their IPT 
> instance? How? Or is there another process that is preferred?
>
> We currently have IPT Version 2.3.6-r3985b6a installed and plan to 
> upgrade to 2.4.0 soon.
>
> A technical answer is what I seek (on behalf of our technical team).
>
> Again my apologies if the answer to my question is easily found and 
> I'm just not finding it.
>
> Annie Simpson, BISON product owner
>
> (she/her/hers)
>
> BioFoundational Data Team
>
> Science Analytics & Synthesis Program
>
> U.S. Geological Survey
>
> 12201 Sunrise Valley Dr. Mailstop 302
>
> Reston VA   20192
>
> asimpson at usgs.gov
>
> +1 703-648-4281
>
> https://orcid.org/0000-0001-8338-5134
>
> https://bison.usgs.gov
>
> Image removed by sender. <https://bison.usgs.gov/>
>
> 	
>
> Biodiversity Information Serving Our Nation (BISON) 
> <https://bison.usgs.gov/>
>
> USGS Biodiversity Information Serving Our Nation (BISON) is a unique, 
> web-based Federal mapping resource for species occurrence data in the 
> United States and its Territories and Canada, including marine 
> Exclusive Economic Zones (EEZs).
>
> bison.usgs.gov
>
>
> _______________________________________________
> IPT mailing list
> IPT at lists.gbif.org
> https://lists.gbif.org/mailman/listinfo/ipt
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.gbif.org/pipermail/ipt/attachments/20200707/2580bd0f/attachment-0001.html>