hibernate oracle sequence produces large gap

I am using hibernate 3 , oracle 10g. I have a table: subject. The definition is here

CREATE TABLE SUBJECT
    ( 
     SUBJECT_ID NUMBER (10), 
     FNAME VARCHAR2(30)  not null, 
     LNAME VARCHAR2(30)  not null, 
     EMAILADR VARCHAR2 (40),
     BIRTHDT  DATE       not null,
     constraint pk_sub primary key(subject_id) USING INDEX TABLESPACE data_index
    ) 
;

when insert a new subject, sub_seq is used to create an subject id, the definition is here

create sequence sub_seq
       MINVALUE 1 
       MAXVALUE 999999999999999999999999999 
       START WITH 1
       INCREMENT BY 1 
       CACHE 100 
       NOCYCLE ;

the Subject class is like this:

@Entity
@Table(name="ktbs.syn_subject")
public class Subject {

    @Id 
    @Column(name="subject_id")
    @GeneratedValue(strategy=GenerationType.SEQUENCE, generator="SUB_SEQ")
    @SequenceGenerator(name="SUB_SEQ", sequenceName = "SUB_SEQ")
    private long subjectId;
    private String fname;
    private String lname;
    private String emailadr;
    private Date birthdt;
}

in the subject table , there have been 4555 subjects in the database loaded by plsql scripts from excel and the sub_sequence worked fine. subject ids range from 1--4555.

however, when i added a subject from my application using hibernate, the sequence number jumped to 255050. After several days running, the subject ids generated by hibernate look like this

270079
270078
270077
270076
270075
270074
270073
270072
270071
270070
270069
270068
270067
270066
270065
270064
270063
270062
270061
270060
270059
270058
270057
270056
270055
270054
270053
270052
270051
270050
265057
265056
265055
265054
265053
265052
265051
265050
260059
260058
260057
260056
260055
260054
260053
260052
260051
260050
255067
255066
255065
255064
255063
255062
255061
255060
255059
255058
255057
255056
255055
255054
255053
255052
255051
255050
4555
4554
4553
.
.
.
.
1

There are several large gaps: 4555 to 255051, 255067 to 260051, 265057 to 270051

this is a waste and not a desired behavior.

does anyone know why this happens and hot to fix it

Thanks


I think that the problem comes from the fact that the sequence generator is not really a sequence generator, but a sequence hilo generator, with a default allocation size of 50. as indicated by the documentation : http://docs.jboss.org/hibernate/stable/annotations/reference/en/html_single/#entity-mapping-identifier

This means that if the sequence value is 5000, the next generated value will be 5000 * 50 = 250000. Add the cache value of the sequence to the equation, and it might explain your huge initial gap.

Check the value of the sequence. It should be less than the last generated identifier. Be careful not to reinitialize the sequence to this last generated value + 1, because the generated valus would grow exponentially (we've had this problem, and had negative integer ids due to overflow)


Agree with JB. But still thanks to PaulJ.

To be more specific to my annotation code below:

@Entity
@Table(name="ktbs.syn_subject")
public class Subject {

  @Id 
  @Column(name="subject_id")
  @GeneratedValue(strategy=GenerationType.SEQUENCE, generator="SUB_SEQ")
  @javax.persistence.SequenceGenerator(name="SUB_SEQ", sequenceName = "SUB_SEQ")
  private long subjectId;
  private String fname;
  private String lname;
  private String emailadr;
  private Date birthdt;
}

If you use javax.persistence.SequenceGenerator, hibernate use hilo and will possibly create large gaps in the sequence. There is a post addressing this problem: https://forum.hibernate.org/viewtopic.php?t=973682

There are two ways to fix this problem

  1. In the SequenceGenerator annotation, add allocationSize = 1, initialValue= 1
  2. instead of using javax.persistence.SequenceGenerator, use org.hibernate.annotations, like this:

    @javax.persistence.SequenceGenerator(
        name = "Question_id_sequence", 
        sequenceName = "S_QUESTION"
    )
    
    @org.hibernate.annotations.GenericGenerator(
        name="Question_id_sequence", 
        strategy = "sequence", 
        parameters = { 
            @Parameter(name="sequence", value="S_QUESTION") 
        }
    )
    

I have tested both ways, which works just fine.


Actually having allocationSize=1 is fine if your sequence INCREMENT VALUE is 1 and you do not have the need of persisting a lot of entities. However if you want to persist thousands or millions of records, the above setting could become a performance bottleneck since every save need to fetch a id hence need a db read.

To Solve this problem, we need to set the allocationSize to something like 500 and sequence INCREMENT VALUE in DB also to 500, then most important add a hibernate setting hibernate.id.new_generator_mappings to ask it to use the new sequence generator implementation, here i assume you set your hibernate properties in a java Config class:

properties.setProperty("hibernate.id.new_generator_mappings", Boolean.toString(true));

This way, Hibernate will use SequenceStyleGenerator rather than the old SequenceHiLoGenerator to generate the ids. The SequenceStyleGenerator is more jpa and oracle friendly. It generates identifier values based on an sequence-style database structure. Variations range from actually using a sequence to using a table to mimic a sequence.

Look at my post for more detail if you are in the same boat:

vcfvct.wordpress.com/2016/04/23/jpa-sequencegenerator-with-allocationsize-1-performance-tuning/


Another solution is:

Use 'GenerationType.AUTO' instead of 'GenerationType.SEQUENCE' as strategy for @GeneratedValue, as below;

@Id
@SequenceGenerator(name = "studentId", sequenceName = "student_Id")
@GeneratedValue(strategy = GenerationType.AUTO, generator="studentId")  
private int studentId;

If you read the following link, you will see that the problem is caused by the CACHE setting on your sequence creation command. Removing the cache setting will solve the problem to a degree - but does not take into account the possiblity of rollbacks, etc.

Link is: http://asktom.oracle.com/pls/apex/f?p=100:11:0::::P11_QUESTION_ID:369390500346406705

The only way to resync your sequences now is to re-create the sequence, rename the current table and create the table again and then re-insert the records from the old table into the new table.

NOTE: The cache value for sequences is useful for large loads where 'x' sequence values are allocated at once. If you are using a transaction system where you do one insert at a time - then caching is not useful (or I should say - I've never found it useful).

NOTE: This is my understanding of the cache option for sequences. You can look up the Oracle Documentation on CREATE SEQUENCE commands for more info. But the link above should provide a reasonable answer to your question.

Thanks. Paul