Hello. Since "increment" is a in-memory generator (hence, not recommended for cluster environments), I'm trying to figure out if hibernate can do a SELECT MAX(id)+1 type generation of the id fields?

Our primary keys aren't setup with SEQUENCE, so I have to do this myself in the app code.

Other than do it manually in the Java code, is there anyway I can have Hibernate do a MAX(id)+1 for me?

I noticed there is a "select" generator, however, that seems to get the data from another column.

... 2 transactions each tries to save a new record at the same time.

they both READ the same MAX(id), and hence will write the same MAX(id)+1 as id for the new objects.

Even if Hibernate can do it, it is just a little bit dodgy?

The problem would be that in your clustered environment, would have to lock your entity table before writing the new row and keep it locked until the new row is inserted. You would have to block both reads and writes, since you are depending on the table's contents to make a new id.

Perhaps the solution would be as simple as making a service that listens on a tcp socket, increments an internal counter by one per tcp connection, and outputs the result to the connecting machine. It would seem that the blocking time doing so would be less then locking the database table for r/w in a cluster while a row is being inserted.

Thanks, guys.

I'm not sure if our production environment will permit a listener type service to generate unique ids.

How about a small/separate database table (say, with two columns, attribute char(10) and nextid int) that can be locked to get a unique id.

I'm also pushing the DBA to allow use of a SEQUENCE column in the database.