Skip to content

Latest commit

 

History

History
159 lines (109 loc) · 6.39 KB

README.md

File metadata and controls

159 lines (109 loc) · 6.39 KB
Raw

islandora-1x-solr-indexer

Information

This replaces the GSearch indexer with a simple camel route that could be extended easily.

Java

This requires Java 11 to build and run

Deployment

  1. Clone the repository.
  2. Change into the directory.
  3. Run ./gradlew build install.
  4. Copy the ./build/libs/islandora-1x-solr-indexer.jar to wherever you'd like.
  5. Copy the example.properties file and edit as necessary.
  6. Run JAR using the environment variable fc3indexer.config.file to point to your file.

Configuration

Configuration is done via a properties file, copy and edit the example.properties file as needed.

Point to your customized properties file using the fc3indexer.config.file variable.

java -Dfc3indexer.config.file=/absolute/path/myproperties.properties -jar islandora-1x-solr-indexer.jar

You MUST configure the location of your XSLT directory in the xslt.path option.

This xslt.path directory should contain XSLT files named with the same name as the datastream ID they will process (ie. RELS-EXT.xslt, DC.xslt, etc)

You MUST have at least one xslt named FOXML.xslt to handle the object XML.

These stylesheets should not output the XML declaration as the resulting XML is re-combined. So please ensure you have a

<xsl:output method="xml" encoding="UTF-8" omit-xml-declaration="yes"/>

in each of your XSLTs.

Besides the xslt.path option the other options are:

jms.brokerUrl=tcp://127.0.0.1:61616
  • The hostname and port of the JMS broker
jms.username=
jms.password=
  • Username/Password (if required) to connect to the JMS broker
queue.incoming=ext-activemq:queue:fedora_update
  • The queue to read incoming messages from. The ext-activemq: part aligns with the JMS bean wired internally to have a single consumer.
queue.internal=activemq:queue:internalIndex

  • The indexer reads messages off of the queue.incoming queue into an aggregator. It will collect all the messages that occur within 10 seconds (configurable with completion.timeout) of each other and only process the last one. That message is passed to this internal queue which specifies activemq: to use the JMS bean which is consumed by the number of consumers defined in jms.processes.

    When ingesting objects in Fedora you normally get a JMS message for each of object ingest, datastream modify, etc. This helps to reduce the redundant indexing.

solr.processes=1
  • Solr update/delete messages are processed this many at a time. Don't overload your Solr box.
fcrepo.baseUri=http://localhost:8080
fcrepo.basePath=/fedora
fcrepo.authUser=fedoraAdmin
fcrepo.authPassword=
  • These define where your Fedora URI and a username/password to allow us to get the datastreams to index.
index.inactive=false
index.deleted=false
  • These allow you to index records that have a status of Inactive and/or Deleted. Normally records without a status of Active are removed from the Solr index, if you enable these options and don't want them displayed you need to add a default filter for only displaying records with the Active status.
solr.baseUrl=solr://localhost:8080/solr

completion.timeout=10000
  • How long (in milliseconds) to wait for messages in the aggregator. Defaults to 10 seconds.
reindexer.port=9111
reindexer.path=/fedora3-solr-indexer
  • On localhost at this port and path a reindexer GET endpoint will be located.
custom.character.file=
  • A file of characters to alter when converting from plain text to XML. If the file exists each line should have the form <character to remove>:<character to replace with>

How it works

This indexer watches the activemq queue for Fedora update messages. When one arrives:

  1. If the header methodName is purgeObject then the object is automatically deleted from Solr using the header pid.
  2. If not, the FOXML is retrieved from Fedora and the property info:fedora/fedora-system:def/model#state is checked. If it is Active the object is indexed, otherwise it is deleted.

Indexing process

The FOXML is split up into foxml:datastream and processed, if the mime-type is text/xml, application/xml, application/rdf+xml, text/html or text/plain the datastream content is retrieved from Fedora and transformed using a stylesheet of the same name (as the datastream ID plus .xslt) in the directory specified by the xslt.path configuration parameter.

If an appropriate XSLT file does not exist, that datastream is skipped.

The datastream ID is available in the XSLT as a parameter called DSID, you can also get the PID with a parameter named pid. These <xsl:param> statements should be at the top level of your XSLTs.

The resulting field XML is concatenated together using the ca.umanitoba.dam.islandora.fc3indexer.utils.StringConcatAggregator wrapped with a <update><doc> </doc><update> and pushed to Solr as an update.

Reindexing

There is a REST endpoint started that allows for forcing a reindex of objects without touching them in Fedora.

Its address is http://localhost:<reindexer.port>/<reindexer.path>/reindex/{pid} where {pid} is the PID of the object to reindex.

It only allows GET requests and responds with a 200 OK and places an item directly onto the queue.internal

Logging/Debugging

If you are experiencing trouble getting your object indexed you can increase the debugging level to TRACE which will give you a tremendous amount of information during processing. It is not recommended to leave the logging at this level for production use.

By default the log level is set to INFO for the indexer and WARN for Camel and other processes (ActiveMQ, Xalan), you can modify the level for the Fedora 3 Indexer or components using the following system properties.

  • fc3indexer.log.indexer = Fedora 3 Indexer
  • fc3indexer.log.camel = Apache Camel
  • fc3indexer.log.activemq = Apache ActiveMQ
  • fc3indexer.log.xml = Xalan and Java.xml

For example to set the indexer to TRACE and Apache Camel to DEBUG

java -Dfc3indexer.log.indexer=TRACE -Dfc3indexer.log.camel=DEBUG -jar islandora-1x-solr-indexer.jar

Credit

All credit to acoburn for this is just an implementation of his camel route.