topic Re: TransactionalCache is full in Alfresco Archive

TransactionalCache is full

mcasanket — Thu, 07 Mar 2013 05:21:31 GMT

Hi All, I have just upgraded alfresco system from 3.2 to 3.4.11. Full indexing took around 24 hours then patches took around 3 hours to be applied. Once the patches were applied I got message that <strong>TransactionalCache</strong> is full and it took more than 12 hours. Because of

Re: TransactionalCache is full

afaust — Mon, 11 Mar 2013 11:28:09 GMT

Hello,

it would be good to find out what kind of logic runs here that is loading all that data into caches. Can you perform some stack dumps using jstack while these messages appear? Have you tried increases logging to DEBUG level to see what kind of services are invoked around those messages? You might also want to adapt the logging format in log4j.properties to include the name of the current thread in logging output.

Without knowing what is going on it is hard to give any advice to overcome this. You could increase cache sizes, but that is only a symptomatic fix.

Regards
Axel

Re: TransactionalCache is full

mcasanket — Wed, 08 May 2013 05:46:22 GMT

Hi AFaust,

<blockquote>it would be good to find out what kind of logic runs here that is loading all that data into caches. Can you perform some stack dumps using jstack while these messages appear? Have you tried increases logging to DEBUG level to see what kind of services are invoked around those messages? You might also want to adapt the logging format in log4j.properties to include the name of the current thread in logging output.</blockquote>

I have tried to connect to JConsole to take the jmx dump but was unable to connect. Also I have set <strong>log4j.logger.org.alfresco.repo.cache.EhCacheTracerJob=debug</strong>. But once the message got printed in log, no other logs come in picture.

It took more than 30 hours to finish up all these cache things. Which is much more than the time it took for full indexing. Please help

Thanks AFaust.

Re: TransactionalCache is full

joe_l3 — Wed, 05 Jun 2013 15:23:00 GMT

Hi,

the WARNING entries show the limit value of the Transactional Cache objects. See the cache-context.xml and ehcache-default.xml files in webapps/alfresco/WEB-INF/classes/alfresco directory.

You can ignore them during the upgrade process.
Probably you have a heavy contentstore and your documents are very large to be cached during reindexing process. I have experience of Alfresco migrations who have involved 6/7 million of documents (about 500 GB of alf_data and 36 hour for Tomcat bootstrap).
During the bootstrap let Alfresco perform the patches
Then I suggest you to make a new migration from 3.4 to 4.0 and to enable the new solr search engine instead of lucene.
Solr engine is fully scalable and you can setup it into another Tomcat Application Server.