Notes about Search Subsystem in Alfresco

Search subsystem in Alfresco can be Lucene or SOLR type. In the
default configuration for Alfresco 4, the search subsystem is SOLR.
This can be set up in alfresco-global.properties

 

index.subsystem.name=solr
solr.port.ssl=8443
solr.host=localhost
dir.keystore=${dir.root}/keystore

   

where the SOLR application is deployed by default on the same
Tomcat container, and it works with the 8443 SSL connector. It is
important, as part of the Alfresco installation, generating the a
certificate browser.p12

 

SOLR application has two cores preconfigured, one for the workspace
and other one for the archive (which is not needed usually). The
configuration can be found in solrcore.properties files:

 

${dir.root}/solr/workspace-SpacesStore/conf/solrcore.properties

${dir.root}/solr/archive-SpacesStore/conf/solrcore.properties

 

They look like:

 

data.dir.root=/opt/alfresco/alf_data/solr
data.dir.store=workspace/SpacesStore
alfresco.host=localhost
alfresco.port=8080
alfresco.port.ssl=8443
alfresco.baseUrl=/alfresco
alfresco.cron=0/15 * * * * ? *
alfresco.stores=workspace://SpacesStore

    

where it is specified, the location for the indices in the disk,
the alfresco repository server (in this case localhost) and a cron job
for tracking the DB indexation count from SOLR app to Alfresco
repository services. In the default installation, SOLR application and
Alfresco repository are installed in the same Tomcat container.

 

• Remember to store your indices in the fastest HD (always).
  Backup indices can be in slower ones. The location is set in conf/solrcore.properties
• Do not forget to generate a new keystore for your Alfresco installation.
• Consider to use SOLR in a different server from Alfresco
  repository (licensed) or at least in another Tomcat container in
  the same machine. This helps to reindex independently of the
  repository. In this last case, reindexing do not mean lost of service.
• Install also an alfresco.war instance in your external SOLR
  server, connected to the Alfresco database for minimizing http
  traffic, in the indexation job.
• If you have a SOLR cluster, consider to dedicate one node for
  indexation and another one for searching. Tune the mergeFactor in
  solrconfig.xml for each core, 25 is ideal for indexing, while 2 is
  ideal for search.
• In solrconfig.xml of each core configure the ramBufferSize to be
  at least 64 Mb.
• In a massive upload of documents consider to disable
  SOLR during the load.
• Other interesting actions of the SOLR admin panel are related to
  get info about the health of SOLR:

• The indices can always regenerate (although it can consume much
  time in large repositories). For SOLR, use this tip, removing the
  indices and reboot SOLR. For Lucene, index.recovery.mode = FULL