cancel
Showing results for 
Search instead for 
Did you mean: 

Alfresco 3.3.5 Enterprise performance lacking

snertz
Champ in-the-making
Champ in-the-making
Hi all,

Currently we have an Alfresco Enterprise production environment, version 3.3.5.
The performance of this environment is lacking since the repository is growing.
If we browse through the Alfresco Explorer it takes up to 10 seconds to open an empty folder. It used to be less then 1 second.

The Alfresco environment is a standard installation with minor content-model changes. We only added a string field to the content metadata-model.

Currently we have the following extra configuration settings apart from the standard settings:

* alfresco-global.properties:

hibernate.jdbc.fetch_size=150
db.pool.max=275
db.pool.validate.query=SELECT 1 FROM DUAL
system.usages.enabled=false

lucene.indexer.maxFieldLength=1000
index.recovery.maximumPoolSize=20
lucene.indexer.maxDocsForInMemoryMerge=20000
lucene.indexer.maxRamInMbForInMemoryMerge=256
lucene.indexer.maxDocsForInMemoryIndex=20000
lucene.indexer.maxRamInMbForInMemoryIndex=256

* setenv.sh
export CATALINA_OPTS="-Xms8G -Xmx8G -XX:MaxPermSize=512m -XXSmiley TongueermSize=256M -XX:NewSize=512M -XX:MaxNewSize=2G -XX:+UseConcMarkSweepGC -XX:+CMSIncrementalMode -XX:+CMSIncrementalPacing -XX:+CMSParallelRemarkEnabled -XX:+UseParNewGC -Xss1m -Dcom.sun.management.jmxremote -Dcom.sun.management.jmxremote.port=5123 -Dcom.sun.management.jmxremote.ssl=false -Dcom.sun.management.jmxremote.authenticate=false -XX:+HeapDumpOnOutOfMemoryError -XX:HeapDumpPath=/opt/alfresco/tomcat/logs"

Our server specifications:

I added two attachments showing the server specifications of the Linux (Alfresco) VM and the Oracle environment.

We tried the following things so far:

- Full reïndexing of the Lucene Indexes.

Do you guys/girls have any idea how we can boost the performance?

Regards,

Vincent Beukers
3 REPLIES 3

mrogers
Star Contributor
Star Contributor
I suggest you call Alfresco Support.
They will almost certainly ask for more information like stack dumps and performance monitoring results.
Why did you think you needed to reindex lucene?   And did it make a difference?

From what you posted db.pool.max=275 may be worth increasing this a bit.     Depends upon your load.
And is your oracle database running O.K. have you gathered the stats recently?

snertz
Champ in-the-making
Champ in-the-making
Hi MrRogers,

The database is running OK, Oracle ADDM is not showing any warnings or errors.
Also we tried to rebuild the Alfresco indexes in the database, this didn't make any difference.

We editted some more config settings which show Alfgresco is faster:

export CATALINA_OPTS="-Xms8G -Xmx8G -XX:MaxPermSize=512M -XXSmiley TongueermSize=512M -server -XX:NewRatio=2 -XX:+UseConcMarkSweepGC -XX:+CMSIncrementalMode -XX:+CMSIncrementalPacing -XX:+CMSParallelRemarkEnabled -XX:+UseParNewGC -Xss1m -Dcom.sun.management.jmxremote -Dcom.sun.management.jmxremote.port=5123 -Dcom.sun.management.jmxremote.ssl=false -Dcom.sun.management.jmxremote.authenticate=false -XX:+HeapDumpOnOutOfMemoryError -XX:HeapDumpPath=/opt/alfresco/tomcat/logs -Djava.awt.headless=true -XX:+CMSClassUnloadingEnabled -XX:+CMSParallelRemarkEnabled -XX:CMSInitiatingOccupancyFraction=80 -XX:ReservedCodeCacheSize=256M"

Can you give your humble opinion about these settings?

Thanks in advance,

Regards,

Vincent

mrogers
Star Contributor
Star Contributor
I still suggest you contact alfresco support.

So how much faster has the application gone with the removal of Max New Size?

In general I think the current advice is to leave the JVM alone without changing properties.    If you have a memory problem then you really need to run a profiler to identify exactly what the problem is.