Can my installation be recovered from a corrupted Lucene index without performing a full reindexing?

Issue

• There is one or more error messages containing information similar to the following lines in one of the cluster nodes log:

  Initializing Lucene writer failed for 155

  java.io.IOException: read past EOF:MMapIndexInput(path="/opt/liferay/data/lucene/155/segments_645o4q")

Environment

• Liferay Portal 6.2

Resolution

• Given that the following properties are enabled:

  cluster.link.enabled=true
  ehcache.cluster.link.replication.enabled=true
  lucene.replicate.write=true

• It would be possible to avoid a full reindexing by forcing to load the master node's indexes into the slave node. These would be the steps:

1. 1. Make a functional backup of the installation in case it were necessary to go back. 
   2. Stop the affected node (only the affected node) . After it stops the remaining active node should be selected as Lucene Master node automatically.
   3. Delete <LIFERAY_HOME>/data/lucene folder in the affected node.
   4. Start the affected node.
   5. Wait until the indexes can be fully loaded from master node:
      • This reindexing may take some considerable time depending of the volume of the current data.
      • Likewise, while the index is being regenerated, some queries may not be available from that node until the reindexing is completed.