AnsweredAssumed Answered

Failed to merge indexes: OutOfMemoryError: Java heap space

Question asked by fschnell on Mar 30, 2007
Latest reply on Mar 30, 2009 by zomurn
Hi
running Alfresco 2.0 Community on a recent HP desktop PC with OpenSuse and 1 GB of physical RAM and 2GB Swap. We are copying a Windows filesystem directory structure into it with the CIFS interface. Approx 400,000 Files totaling ~120 GB to evaluate performance und usability aspects. After 50,000 files we receive error messages in the Alfresco.log file:

10:45:01,827 INFO  [org.alfresco.repo.module.ModuleServiceImpl] Found 0 module(s).
10:45:02,315 INFO  [org.alfresco.service.descriptor.DescriptorService] Alfresco JVM - v1.5.0_08-b03; maximum heap size 506.313MB
10:45:02,316 INFO  [org.alfresco.service.descriptor.DescriptorService] Alfresco started (Community Network): Current version 2.0.0 (build-185) schema 38 - Installed version 2.0.0 (build-185) schema 38
11:35:23,444 ERROR [org.alfresco.repo.search.impl.lucene.index.IndexInfo] Failed to merge indexes
java.lang.OutOfMemoryError: Java heap space
11:35:59,246 ERROR [org.alfresco.repo.search.impl.lucene.index.IndexInfo] Failed to merge indexes
java.lang.OutOfMemoryError: Java heap space
11:38:27,716 ERROR [org.alfresco.smb.protocol] Closing session due to throwable
java.lang.OutOfMemoryError: Java heap space
11:38:28,829 ERROR [org.alfresco.smb.protocol] Socket closed by remote client
11:40:01,848 ERROR [org.alfresco.repo.search.impl.lucene.index.IndexInfo] Failed to merge indexes
java.lang.OutOfMemoryError: Java heap space
11:40:53,175 ERROR [org.alfresco.util.transaction.SpringAwareUserTransaction] Transaction didn't commit
java.lang.OutOfMemoryError: Java heap space
12:57:08,117 ERROR [org.alfresco.smb.protocol] Socket closed by remote client
13:50:24,037 ERROR [org.alfresco.smb.protocol] Socket closed by remote client

At the same time 'top' reports 1 GB of RAM used. The copy process stopped. Despite that and no user using it, the RAM does not get released. Only a stop-start Tomcat will work, until the copy process dies again.

How much memory should we foresee for the indexer to work properly with very many files?

Frank

Outcomes