Hibernate Seach lock the sesion

markosca · **Joined:** Mon Feb 06, 2012 10:53 am **Posts:** 9

Hello!

I'm doing an a Web application with struts and hibernate and all works good, but now i want add hibernate search.
I have a problem because i'm using HibernateUtils to generate a session but when i add hibernate Search and I start the application. In the log don't appear any error but when the application wants to get a sesion, apparently, the application is blocked.
Debuging this case I can see that the code stops when each dao of application try to get the sesion, because (i think) hibernate seach has this sesion.

Some idea?
Thanks and sorry for my english.

markosca · **Joined:** Mon Feb 06, 2012 10:53 am **Posts:** 9

It was the size of the pool. I'm using org.hibernate.connection.C3P0ConnectionProvider and if I set a size less than 15, the indexer doesn't end the process.

is there any form to start correctly the application with a less pool size?

Thanks !

sanne.grinovero · **Posted:** Wed Feb 08, 2012 12:48 pm

Hi,
Hibernate Search should not use any additional connection. The only known case is when you have it use the MassIndexer, in that case it will need a lot of connections, depending on how many threads you have started it with and how many indexed types you have.

https://hibernate.onjira.com/browse/HSEARCH-598

markosca · **Joined:** Mon Feb 06, 2012 10:53 am **Posts:** 9

Hi,

Hibernate doesn't use any additional connection, but I'm using the MassIndexer and it needs a big pool, but I haven't so connections.
If this size of pool size is mandatory, this will be a problem for me, because I only have 10 connections.

I'm trying to configure the MassIndexer when it starts to see if with a configuration is able to start with a pool size of 10, but I don't find it.

Code:

   massIndex.purgeAllOnStart(true) // true by default, highly recommended
            .threadsForIndexWriter(7)
            .batchSizeToLoadObjects( 25 )
            .threadsToLoadObjects( 5 )
            .threadsForSubsequentFetching( 10 )
            .startAndWait();

sanne.grinovero · **Posted:** Thu Feb 09, 2012 6:27 am

The issue I mentioned in previous post is just meant to make this configuration easier, but it's possible even with current version.

Here is a workaround:

threadsForSubsequentFetching + threadsToLoadObjects + 1 = number of connections needed for each indexed entity

So you have to
1) pick a lower amount of threads especially for the two options mentioned above.
2) Index one entity type at a time.

Code:

fullTextSession.createIndexer( EntityA.class ).batchSizeToLoadObjects( 25 ).threadsToLoadObjects( 3 ).threadsForSubsequentFetching( 5 ).threadsForIndexWriter( 2 ).startAndWait();
fullTextSession.createIndexer( EntityB.class ).batchSizeToLoadObjects( 25 ).threadsToLoadObjects( 3 ).threadsForSubsequentFetching( 5 ).threadsForIndexWriter( 2 ).startAndWait();
//... other entities, doing it in series instead of in parallel
 

markosca · **Joined:** Mon Feb 06, 2012 10:53 am **Posts:** 9

Hi,

I have just tried and works fine.

Thanks so much for your answer.