Hello,

Is there any limitation in using Hibernate with Oracle 9i RAC ?

The way we are willing to use Hibernate is through JBoss DataSources and Oracle OCI client.

Thanks for any feed-back

BB

Hibernate is just like any other JDBC client, unless you enable the second-level cache, in which case it is just like any other JDBC client that has a cache.

Please do not crosspost the forums, posting once is enough. I have answered in your other post here http://forum.hibernate.org/viewtopic.php?t=927502

I am not a developer, so please take it easy on me. We are implementing an Oracle 10.1.0.4 RAC cluster in the same timeframe that our application counterparts are moving to Hibernate from a home-grown solution.

The application teams involved would like to know if there is any special configuration or customization that will be required in order for Oracle connection pools managed by Hibernate to work properly during a failure. The connection string will handle load balancing, etc., but I assume that things like detecting failures, reaping dead connections, reconnecting (in a loop) after a failure, etc. would be managed by Hibernate. Is that true?

You can probably tell that I don't know exactly what I should be asking, but if anyone has experience using Hibernate and Oracle 10g RAC together, I would be interested in hearing about any problems or issues that have been identified.

Thanks in advance for your help!

From the application's standpoint, AFAIK the JDBC driver abstracts all RAC aspects away from the application. The Java code could care less whether or not it is talking to one Oracle node or n nodes. I've been using this for months now with no application-level issues at the JDBC level. You'll just need to create a connect string that lists both nodes using the funky Oracle nested parenthesis format.

A variation of this works fine for me against 9iR2 RAC. I would think 10g works the same.



On the flip side, the DBA shouldn't be too concerned about the SQL Hibernate generates. Mostly its just doing "keyed reads" against the database tables. You can enable logging for all SQL Hibernate generates, so if you want to check certain statements for performance, you'll have access to all database interactions. I recommend using something that will autoformat SQL (e.g. Toad).

By the way, does anyone know of any good SQL formatters that are free?

We currently have an Oracle 9i database and are primarily using DAO's to access the database using connection pools set up on our Websphere application servers. We've recently started switching our strategy to use Hibernate for all the benefits it provides and have already switched many of our objects to use this.

Now, due to the need to scale out, we are planning on going to 10g and implementing RAC. I'm just coming up to speed on this as well, but I know there are code changes required around inserts, updates and deletes in order to catch errors received by one of the nodes going down and retry those queries on another node.

My question is similar to sandeaj's in that I want to know if Hibernate can be configured to know about RAC and automatically take care of wrapping the SQL to be aware of the error and retry the statement without any manual changes by the developer. If not, is anyone come up with a good way to implement automatic transaction failover with Hibernate in a generic way? Samples would be helpful.

Thanks!

I think that the transparent application failover (TAF) only works with the OCI drivers--keep that in mind. There's two different levels you can adopt, if I understand correctly (although I may have this wrong). One level is session failover, where the connection won't be dropped if one node shuts down. Any open cursors are terminated along with the transaction, but the connection remains open. The other type is cursor level, which lets you resume an existing select statement.

It appears that the JDBC driver handles TAF for you, if possible. Take a look at the sample app at:

http://www.oracle.com/technology/sample ... eadme.html

The QueryThread.java class looks like it is just using plain-ol JDBC. Since Hibernate just uses plain JDBC, I'm assuming TAF would work fine with Hibernate.

Keep in mind that this is not a silver bullet--your uncommitted inserts/updates/deletes will roll back if the node goes down. See:

I hope this helps. Post your findings to this thread so others can benefit. Likewise, I'll give this a try on my cluster.

It would help if I pasted the other URL :-)

http://www.remote-dba.cc/10g_10_1.htm