Harvesting & Importing

Note

All of these commands should be run within the catalog or cron Docker container

Script Summary

pc-import-folio: Used to harvest and import FOLIO MARC data into the biblio collection of Solr.
pc-import-hlm: Used to harvest and import EBSCO MARC data into the biblio collection of Solr from the FTP location given access to by EBSCO. The records contain the HLM dataset that is missing from FOLIO's database.
pc-import-authority: Used to harvest and import MARC data from Backstage into the authority collection in Solr from the FTP location provided by Backstage.
cron-reserves.sh:
pc-full-import: Wrapper script for the pc-import-folio and pc-import-hlm scripts to perform a full import of data.

HLM Data

One of our sources is HLM (Holdings Link Management) which is files from EBSCO's FTP server with electronic resources our library has access to which are not in FOLIO. To automate the retrieval and import of these files we have a wrapper script, pc-import-hlm.

Since occasionally EBSCO sends us full sets of data, we want to be able to exclude all previous sets from them, so to do this we have an ignore files /mnt/shared/oai/ignore_patterns.txt with will check to see if the file name contains any of those substrings and ignore them if they match.

Also, we will only harvest files that match the pattern *.m*c and *.zip. *.m*c files will be imported with the exception of -del*.m*c (case insensitive) or .delete files which will instead be tagged as deletion files.

If you simply want to see what files are on the FTP server currently, you can run our helper script, pc-list-hlm-remote, to list all the files. If you want to download a specific file, run pc-get-hlm-remote [NAME OF FILE].

Full Data Harvests

This section describes the steps needed to re-harvest all the data from each source.

FOLIO

Ensure that your OAI settings on the FOLIO tenant are what you want them to be for this particular harvest. For example, if you wish to include storage and inventory records (i.e. the records without a MARC source) then you will need to modify the "Record Source" field in the OAI Settings in FOLIO.
Next you will need to clear out the contents of the harvest_folio directory before the next cron job will run. Assuming you want to preserve the last harvest for the time being, you can simply move those directories somewhere else and rename them. Below is an example, but certainly not the only option. The only goal is that the harvest_folio directory has no files in it, but can have the log and processed directories within it as long as they are empty (they technically can have files in them, you just will not want them to have files since they will get mixed in with your new harvest).
```
cd /mnt/shared/oai/[STACK_NAME]/harvest_folio/
mv processed processed_old
mv log log_old
mv last_state.txt last_state.txt.old
mv harvest.log harvest.log.old
```
Monitor progress after it starts via the cron job in the monitoring app or in the log file on the container or volume (/mnt/logs/harvests/).

HLM

This can just be done with the script's --full --harvest flags in a one-off run, but if you prefer to have it run via the cron job use it's normal flags, here are the steps you would need to do in order to prepare the environment.

Remove all files from the /mnt/shared/hlm/[STACK_NAME]/current/ directory and remove all files from the container's local/harvest/hlm. You can also just move them somewhere else if you want to preserve a copy of them.

find /mnt/shared/hlm/[STACK_NAME]/current/ -maxdepth 1 -name '*.marc' -print0 | tar -czf archive_[SOME_DATE].tar.gz --null -T -
find /mnt/shared/hlm/[STACK_NAME]/current/ -maxdepth 1 -name '*.marc' -delete

# exec in to the container and run
find /usr/local/vufind/local/harvest/hlm -mindepth 1 -maxdepth 1 -name '*.marc' -delete
rm /usr/local/vufind/local/harvest/hlm/processed/*

Monitor progress after it starts via the cron job in the monitoring app or in the log file on the container or volume (/mnt/logs/harvests/).

Backstage (Authority records)

This can just be done with the script's --full --harvest flags in a one-off run, but if you prefer to have it run via the cron job use it's normal flags, here are the steps you would need to do in order to prepare the environment.

Remove all files from the /mnt/shared/authority/[STACK_NAME]/current/ directory and remove all files from the container's local/authority/hlm. You can also just move them somewhere else if you want to preserve a copy of them.

tar -czf archive_[SOME_DATE].tar.gz /mnt/shared/authority/[STACK_NAME]/current/
rm /mnt/shared/authority/[STACK_NAME]/current/processed/*
rm /mnt/shared/authority/[STACK_NAME]/current/*

# exec in to the container and run
rm /usr/local/vufind/local/harvest/authority/*
rm /usr/local/vufind/local/harvest/authority/processed/*

Monitor progress after it starts via the cron job in the monitoring app or in the log file on the container or volume (/mnt/logs/harvests/).

Full Data Imports

`biblio` Index

Helper script for full import

There is now a helper script to run all of the below steps in the proper order to do a full re-import of data in the biblio index. See the full documentation for the pc-full-import and the below for a simple example.

sudo screen
# Prompting for confirmation
pc-full-import catalog-prod --debug 2>&1 | tee /mnt/shared/logs/catalog-prod-import_$(date -I).log
# Bypassing user confirmation
pc-full-import catalog-prod --yes --debug 2>&1 | tee /mnt/shared/logs/catalog-prod-import_$(date -I).log

Should you choose to do the steps manually, this section will describe the process needed to run a full re-import of the data since that is frequently required to update the Solr index with new field updates. If other tasks are required (such as full harvests or incremental) refer to the --help flags on the appropriate script.

Full imports for the biblio collection can be done

directly in the cron container for prod/beta/preview,
in the catalog container for dev environments,
using the biblio collection alias in the build container, to avoid serving incomplete collections in prod (see How to Use the Collection Aliases to Rebuild and Swap below).

Importing FOLIO records using the cron container

Connect to one of the catalog server nodes and move the following files up a level out of the processed directory in the shared storage. This will allow them to be picked up by the next cron job and re-started automatically should the container get stopped due to deployments. Progress can be monitored by checking then number of files remaining in the directory and the log file in /mnt/logs/harvests/folio_latest.log.

# Can be done inside or outside the container
mv /mnt/shared/oai/[STACK_NAME]/harvest_folio/processed/* /mnt/shared/oai/[STACK_NAME]/harvest_folio/

Importing FOLIO records in dev environments

This will import the tests records. In the catalog container:

./pc-import-folio -c /mnt/shared/oai/devel-batch -l 1 -b -v -r

Importing HLM records using the cron container

Assuming HLM records also need to be updated in the biblio index as well, you will need to copy those files from the shared directory into the container prior to starting the script. Then start a screen session and connect to the container again and run the command to import the files. Note that this will get terminated and not be recoverable if the container stops due to a deploy like the previous command was. Process can be monitored by seeing the remaining files in the /usr/local/harvest/hlm/ directory and by re-attaching to the screen (by using screen -r) to see if the command has completed.

# Done inside the catalog_cron container
cp /mnt/shared/hlm/[STACK_NAME]/current/* /usr/local/vufind/local/harvest/hlm/

# You will want to kick off this command in a screen session,
# since it can take many hours to run
/usr/local/bin/pc-import-hlm -i -v

How to Use the Collection Aliases to Rebuild and Swap

As mentioned in the Solr documentation, biblio uses aliases to manage directing VuFind to the collection in Solr that have the "live" biblio data that should be used for searching: biblio1 or biblio2. This means we will on occasion need to swap them. This occasion being when we rebuild the index, such as when we're adding new data fields or doing a VuFind version upgrade (...which typically add new data fields).

Start the manual task "Deploy VuFind Build Env" in GitLab. It will update the catalog_build container. This is not done automatically so that other updates to the main branch can be deployed while a full import is running.
Identify what collection each alias is pointing to currently (i.e. is biblio pointing to biblio1 or biblio2) and confirm the other collection is what biblio-build is pointing to. To get the list of aliases, from a container:

curl -s "http://solr:8983/solr/admin/collections?action=LISTALIASES" | grep biblio

Rebuild the index on biblio-build using the catalog_build container. This has everything that the catalog_cron containers have access to, but do not run cron jobs since rebuilds do not happen at regular or frequent intervals. In fact, all this container does is sleep! It is recommended to run these commands in a screen.

Warning

If you run a deploy pipeline while this is running, you will not want to run the manual job that deploys the updates to the build container (since not all the import scripts are configured to resume where they left off yet).

# On Host
screen
docker exec -it $(docker ps -q -f name=catalog-prod-catalog_build) bash

# Inside container
rm local/harvest/folio/processed/*
cp /mnt/shared/oai/${STACK_NAME}/harvest_folio/processed/* local/harvest/folio/
/usr/local/bin/pc-import-folio --verbose --reset-solr --collection biblio-build --batch-import | tee /mnt/shared/logs/folio_import_${STACK_NAME}_$(date -I).log
[Ctrl-a d]

# On Host
screen
docker exec -it $(docker ps -q -f name=catalog-prod-catalog_build) bash

# Inside container
cp /mnt/shared/hlm/${STACK_NAME}/current/* local/harvest/hlm/
/usr/local/bin/pc-import-hlm --import --verbose | tee /mnt/shared/logs/hlm_import_${STACK_NAME}_$(date -I).log
[Ctrl-a d]

Verify the counts are what you expect on the biblio-build collection using the following command

curl 'http://solr:8983/solr/admin/metrics?nodes=solr1:8983_solr,solr2:8983_solr,solr3:8983_solr&prefix=SEARCHER.searcher.numDocs,SEARCHER.searcher.deletedDocs&wt=json'

Build the spellchecking indices

Building these indices is only necessary for a full import.

curl 'http://solr1:8983/solr/biblio-build/select?q=*:*&spellcheck=true&spellcheck.build=true' &
curl 'http://solr2:8983/solr/biblio-build/select?q=*:*&spellcheck=true&spellcheck.build=true' &
curl 'http://solr3:8983/solr/biblio-build/select?q=*:*&spellcheck.true&spellcheck.build=true' &
wait
curl 'http://solr1:8983/solr/biblio-build/select?q=*:*&spellcheck.dictionary=basicSpell&spellcheck=true&spellcheck.build=true' &
curl 'http://solr2:8983/solr/biblio-build/select?q=*:*&spellcheck.dictionary=basicSpell&spellcheck=true&spellcheck.build=true' &
curl 'http://solr3:8983/solr/biblio-build/select?q=*:*&spellcheck.dictionary=basicSpell&spellcheck=true&spellcheck.build=true' &
wait

/bitnami/solr/server/solr/biblioN/spellShingle and /bitnami/solr/server/solr/biblioN/spellchecker should have a significant size afterward in the solr container (replace biblioN by the biblio-build collection)

Your Solr instance will likely require more memory than it typically needs to do the collection alias swap. Be sure to increase and deploy the stack with additional SOLR_JAVA_MEM as required to ensure no downtime during this step. Currently, 6G (which we use in prod) is enough for the swap. Alternatively (for beta and preview), let it crash after these commands and restart the pipeline to help Solr cloud fix itself.

# Open the solr-cloud compose file for your environment
vim docker-compose.solr-cloud.yml

# Modify the memory line to:
SOLR_JAVA_MEM: -Xms8192m -Xmx8192m

# Now on the host, run the deploy helper script
sudo pc-deploy [ENV_NAME] solr-cloud

Once you are confident in the new data, you are ready to do the swap! BE SURE TO SWAP THE NAME AND COLLECTION IN THE BELOW COMMAND EXAMPLE

Warning

# Command to check the aliases (repeated from above)
curl -s "http://solr:8983/solr/admin/collections?action=LISTALIASES" | grep biblio

# This EXAMPLE sets biblio-build to biblio2, and biblio to biblio1
# biblio-build => biblio2
# biblio       => biblio1
curl 'http://solr:8983/solr/admin/collections?action=CREATEALIAS&name=biblio-build&collections=biblio2'
curl 'http://solr:8983/solr/admin/collections?action=CREATEALIAS&name=biblio&collections=biblio1'

# This EXAMPLE sets biblio-build to biblio1, and biblio to biblio2
# biblio-build => biblio1
# biblio       => biblio2
curl 'http://solr:8983/solr/admin/collections?action=CREATEALIAS&name=biblio-build&collections=biblio1'
curl 'http://solr:8983/solr/admin/collections?action=CREATEALIAS&name=biblio&collections=biblio2'

If needed, back-date the timestamp on your last_harvest.txt re-harvest some of the OAI changes since you started the import
Clear out the collection that biblio-build is pointing to, to avoid having two large indexing stored for a long period of time (only after you are confident in the new index's data)

/usr/local/bin/pc-import-folio --verbose --reset-solr --collection biblio-build

If SOLR_JAVA_MEM was increased, lower it to its previous amount.
Kick off a manual alpha browse re-index if you don't want it to be outdated until the next scheduled run.

# Run this on all of the host's
docker exec -it \
  $(docker ps -q -f name=${STACK_NAME}-solr_cron) \
  /alpha-browse.sh -v -f

`authority` Index

Similar to the process for HLM records, copy the files from the shared directory into the containers import location prior to starting the script. Then start a screen session and connect to the container again and run the command to import the files. Note that this will get terminated and not be in a recoverable state if the container is stopped due to a deploy. Process can be monitored by seeing the remaining files in the /usr/local/harvest/authority directory and by re-attaching to the screen (by using screen -r) to see if the command has completed.

# Done inside the container
cp /mnt/shared/authority/[STACK_NAME]/current/processed/*.xml /usr/local/vufind/local/harvest/authority/

# You will want to kick off this command in a screen session,
# since it can take many hours to run
/usr/local/bin/pc-import-authority -i -B

`reserves` Index

The course reserves data is refreshed by a Cron job on a nightly basis, so likely you will not need to run this manually if you can just wait for the regular run. But if needed, here is the command to run it off-schedule.

# Done inside the container (ideally within a screen since it will
# take hours to run)
php /usr/local/vufind/util/index_reserves.php

Alternatively, you can also modify the cron entry (or add a temporary additional cron entry) in the cron container for the cron-reserves.sh command to run at an earlier time. The benefit of this would be it would save logs to /mnt/logs/vufind/reserves_latest.log and track them in the Monitoring site.

Adding generated call numbers

This should be done after each full import (FOLIO + HLM) when data was reset, in the solr_solr container:

python3 ./add_generated_call_numbers.py

Partial call numbers are added to callnumber-label for records that didn't have any when the call_numbers.csv file was generated.

Note that the call numbers in /mnt/shared/call-numbers/call_numbers.csv are meant for beta/preview/prod. There is another file at /mnt/shared/call-numbers/test_call_numbers.csv that can be used for testing in dev.

Ignoring certain HLM files

If your EBSCO FTP server is set up in a way where it contains all of the sets ever generated for you, then you'll likely want a way to have the pc-import-hlm script ignore the past sets assuming you get new full sets periodically. This can be done by adding a new substring pattern to ignore to the top level of the hlm directory in the shared storage (/mnt/shared/hlm/ignore_patterns.txt). This file is used automatically and created on the cron containers startup if it doesn't exist. You can override the file path by using the -p|--ignore-file flag.

Using VuFind Utilities

The preferred method is to use the included wrapper script with this repository. The pc-import-folio script can run either, or both, the harvest and import of data from FOLIO to VuFind. Use the --help flag to get information on how to run that script.

But should you choose to run the commands included directly with VuFind, below is documentation on how to do that.

Harvesting from Folio

cd /usr/local/vufind/harvest
php harvest_oai.php

## This step is optional, it will combine the xml files into a single file
## to improve the speed of the next import step.
find *.xml | xargs xml_grep --wrap collection --cond "marc:record" > combined.xml
mkdir unmerged
mv *oai*.xml unmerged/

Importing into VuFind

cd /usr/local/vufind/harvest
./batch-import-marc.sh folio