I have just created my first EJB3 Entity bean using the oracle seqeunce. Every time i start the app server and insert a new record the id generated is incremented by 50 instead of using the seqeunce value from the database. I have noticed that the stategy being used is org.hibernate.id.SequenceHiLoGenerator. Is there any way of changing this to use the database sequence value instead of one provided by the app server or the entity manager.

After I switched to annotations from having mapping information in hibernate.cfg.xml, Hibernate isn't using the sequence generated number for the id. I followed the example in Chapter 1 of the Hibernate Reference Documentation, and when the mapping for the Event class was in hibernate.cfg.xml it worked just fine. As soon as I switched to annotations the objects get stored in the database with ids completely different from what are being generated by the sequence, and they increment by 50, as the previous poster experienced. I'm using hibernate-3.2 and hibernate-annotations-3.2.0.CR1. Here's the annotated class:



This is what I notice in the log:



Notice that HIbernate is getting the right sequence number (41), but then is using SequenceHiLoGenerator, which makes the id 2050. If I run the example again the sequence will generate 42 and the id will be 2100. Any ideas on what I might be doing incorrectly?

I have posted this on the EJB3 forum and the response i received was to set the allocationSize=1 in the @SequenceGenerator(name="ID_SEQ",sequenceName="IDSEQ",allocationSize=1)

That fixed it. Thanks.

Is this bug going to be fixed?

I can't understand why it doesn't just pull the value from the Oracle sequence. When we reviewed the logs after getting the same error, it appeared to actually select the value from the sequence, then this AbstractSaveEventListener class picked a random value using strategy: org.hibernate.id.SequenceHiLoGenerator. Any information on why it behaves this way would be appreciated.

Drew

I have been informed that this is not a bug but the behavior. The app server will set the sequence number to +50 every time the server is started. This is to avoid trips to the database for a sequence number.

use @µSequenceGenerator(increment=1) if you don't want the default behavior

SequenceHiLoGenerator is the default sequence generator for JPA. Some guys reported this in Hibernate JIRA and it was declared not a bug, that is, SequenceHiLoGenerator is the default by design and will stay like this: http://opensource.atlassian.com/project ... se/ANN-354

Why has it been decided so?

Probably because it offers performance gains in large scale distributed environments.

Using the hi/lo algorithm allows the application on the given node (in a cluster) to avoid communicating with the database each time it needs a new identifier - depending on allocationSize.

For example, with the default allocationSize value of 50, it needs to select a next value from the sequence only every 50th generated identifier.

It works like this:

1) Gets a "hi" value from the database sequence
2) Multiplies the "hi" value by allocationSize (50 by default)
3) Uses the resulting value as the first generated identifier.
4) When a new identifier is needed, increments the value by 1 without contacting the database, until the pool of 50 available values is depleted. Note that these values are guaranteed to be unique because other nodes in the cluster will reserve other number pools, since they've multiplied other database sequence values by 50.
5) When all 50 values have been used up, it's time to contact the database for a new sequence value that reserves another pool of 50 numbers

Of course, this clever scheme works only as a gentlemen's agreement between all application nodes in the cluster. If any of those starts using a different allocationSize value, the calculated identifiers will sooner or later start colliding.

So the most foolproof behaviour is with allocationSize = 1, which goes back to getting a sequence number from database for each identifier.

BTW, I've reported a JIRA on clarifying this in the JavaDOCs:

http://opensource.atlassian.com/project ... e/HHH-4871

Did you mean allocationSize instead of increment? I can't find any annotation that uses increment....