cgspace-notes/content/posts/2018-07.md

title	date	author	tags
July, 2018	2018-07-01T12:56:54+03:00	Alan Orth	Notes

2018-07-01

• I want to upgrade DSpace Test to DSpace 5.8 so I took a backup of its current database just in case:

$ pg_dump -b -v -o --format=custom -U dspace -f dspace-2018-07-01.backup dspace

• During the mvn package stage on the 5.8 branch I kept getting issues with java running out of memory:

There is insufficient memory for the Java Runtime Environment to continue.

• As the machine only has 8GB of RAM, I reduced the Tomcat memory heap from 5120m to 4096m so I could try to allocate more to the build process:

$ export JAVA_OPTS="-Dfile.encoding=UTF-8 -Xmx1024m"
$ mvn -U -Dmirage2.on=true -Dmirage2.deps.included=false -Denv=dspacetest.cgiar.org -P \!dspace-lni,\!dspace-rdf,\!dspace-sword,\!dspace-swordv2 clean package

• Then I stopped the Tomcat 7 service, ran the ant update, and manually ran the old and ignored SQL migrations:

$ sudo su - postgres
$ psql dspace
...
dspace=# begin;
BEGIN
dspace=# \i Atmire-DSpace-5.8-Schema-Migration.sql
DELETE 0
UPDATE 1
DELETE 1
dspace=# commit
dspace=# \q
$ exit
$ dspace database migrate ignored

• After that I started Tomcat 7 and DSpace seems to be working, now I need to tell our colleagues to try stuff and report issues they have

2018-07-02

• Discuss AgriKnowledge including our Handle identifier on their harvested items from CGSpace
• They seem to be only interested in Gates-funded outputs, for example: https://www.agriknowledge.org/files/tm70mv21t

2018-07-03

• Finally finish with the CIFOR Archive records (a total of 2448):
  • I mapped the 50 items that were duplicates from elsewhere in CGSpace into CIFOR Archive
  • I did one last check of the remaining 2398 items and found eight who have a cg.identifier.doi that links to some URL other than a DOI so I moved those to cg.identifier.url and cg.identifier.googleurl as appropriate
  • Also, thirteen items had a DOI in their citation, but did not have a cg.identifier.doi field, so I added those
  • Then I imported those 2398 items in two batches (to deal with memory issues):

$ export JAVA_OPTS="-Dfile.encoding=UTF-8 -Xmx1024m"
$ dspace metadata-import -e aorth@mjanja.ch -f /tmp/2018-06-27-New-CIFOR-Archive.csv
$ dspace metadata-import -e aorth@mjanja.ch -f /tmp/2018-06-27-New-CIFOR-Archive2.csv

• I noticed there are many items that use HTTP instead of HTTPS for their Google Books URL, and some missing HTTP entirely:

dspace=# select count(*) from metadatavalue where resource_type_id=2 and metadata_field_id=222 and text_value like 'http://books.google.%';
 count
-------
   785
dspace=# select count(*) from metadatavalue where resource_type_id=2 and metadata_field_id=222 and text_value ~ '^books\.google\..*';
 count
-------
     4

• I think I should fix that as well as some other garbage values like "test" and "dspace.ilri.org" etc:

dspace=# begin;
dspace=# update metadatavalue set text_value = regexp_replace(text_value, 'http://books.google', 'https://books.google') where resource_type_id=2 and metadata_field_id=222 and text_value like 'http://books.google.%';
UPDATE 785
dspace=# update metadatavalue set text_value = regexp_replace(text_value, 'books.google', 'https://books.google') where resource_type_id=2 and metadata_field_id=222 and text_value ~ '^books\.google\..*';
UPDATE 4
dspace=# update metadatavalue set text_value='https://books.google.com/books?id=meF1CLdPSF4C' where resource_type_id=2 and metadata_field_id=222 and text_value='meF1CLdPSF4C';
UPDATE 1
dspace=# delete from metadatavalue where resource_type_id=2 and metadata_field_id=222 and metadata_value_id in (2299312, 10684, 10700, 996403);
DELETE 4
dspace=# commit;