RAC and JBOSS Thin client failover
Date: Mon, 19 Oct 2009 13:35:03 -0500
Message-ID: <85C671920FB8CA46B50E423346685B5006AB5C56_at_s228130hz1ew08.apptix-01.savvis.net>
We are running 10.2.0.4 RH4 2 node RAC cluster with ASM. For a middle tier we are using JBOSS hibernate.
I understand we cannot use TAF for failover b/c JDBC thin does not support it. I would like to get as close to that functionality as possible, without having to make code changes to existing programs. As I understand it, I have 2 options:
Based on Metalink DOC 433827.1:
- FCF. This could be done in a number of ways, but to avoid coding changes, it seems using ONS on the middle tier and Implicit Connection Caching will allow us to make failover to the other instance without code changes. This is because once ONS notifies ICC it will invalidate connects to the downed instance, and force Java to reconnect (this assumes the connection to both instances is in the data source).
Based on Metalink DOC 213412.1:
2) Connect time failover. From what I can tell this looks like putting the same entry you would put in tnsnames.ora into your jdbc thin client. If TAF does not work with jdbc thin and there are no coding changes necessary, does this mean after the connection drops, Java will just try to reconnect to the next instance in the data source? So the difference between TAF and connect time failover is that TAF will sometimes transparently failover a select statement, and connect time failover will not?
Is anybody using either of the above? Any feedback is appreciated.
Thanks,
Mike
-- http://www.freelists.org/webpage/oracle-lReceived on Mon Oct 19 2009 - 13:35:03 CDT