Google Groups no longer supports new Usenet posts or subscriptions. Historical content remains viewable.
Dismiss

Websphere 5/Oracle RAC Failver issue

2 views
Skip to first unread message

walterjg

unread,
Oct 6, 2005, 11:15:49 AM10/6/05
to
We are running Websphere 5.1 app from IBM zOS platform that is
connecting to an Oracle RAC
database using the 10.2 "thin" JDBC driver (ojdbc14.jar). For database
failover we implemented connect-time failover in the JDBC URL. The URL
format that
we're using is:

jdbc:oracle:thin:@(DESCRIPTION=(ADDRESS_LIST=(ADDRESS=(PROTO
=TCP)(PORT=1521)
HOST=hostname1))(ADDRESS=(PROTO
=TCP)(PORT=1521)(HOST=hostname2)))(CONNECT_DATA
(SERVICE_NAME=DBNAME.DOMAIN)))

When we bring the primary database instance down, the app does not
connect to the backup instance on the next attempt. Instead it
generates an exception that includes ORA-12514 error:
Io exception: Connection
refused(DESCRIPTION=(TMP=)(VSNNUM=153093632)(ERR=12514)(ERROR_STACK=(ERROR=(CODE=12514)(EMFI=4))))

We cannot re-create this issue outside of Websphere 5.1. The database
failover works fine outside of Websphere so this appears to be a WAS
connection pooling problem. IBM support is looking into this but has
offered no assistance to date.

Has anyone else seen this?
TIA

Rodrick Brown

unread,
Oct 6, 2005, 8:37:29 PM10/6/05
to

"walterjg" <greg_w...@yahoo.com> wrote in message
news:1128611749.8...@o13g2000cwo.googlegroups.com...

JDBC is a highly stateful protocol and in which transactional state are tied
directly to the socket between the server and the client, so it is hard to
achieve clustering. If a JDBC connection dies, a ll JDBC objects associated
with the dead connection will also be defunct. The re-connection action is
needed in the client code. JDBC multipool can ease the reconnection
process, but not all the time.

--
Rodrick R. Brown
Unix Systems Architect
http://www.rodrickbrown.com


walterjg

unread,
Oct 6, 2005, 10:48:57 PM10/6/05
to
Rodrick, thanks for the response. From what you're saying it sounds
like all existing connections within the WAS connection pool are
probably defunct after the primary DB instance crashes. If we can
somehow flush the pool and establish a new pool we may be able to get
around this problem? Is there a way to do this in WAS 5? Any ideas
would be appreciated. The interesting thing here is that is that this
problem appears to be resolved in WAS 6. Unfortunately we are just
starting to test WAS 6.

Rodrick Brown

unread,
Oct 7, 2005, 8:25:51 AM10/7/05
to

"walterjg" <greg_w...@yahoo.com> wrote in message
news:1128653337.2...@g47g2000cwa.googlegroups.com...

Yes you are correct WAS 6 does address this problem.

--
RB


solomanb

unread,
Oct 24, 2005, 9:24:00 AM10/24/05
to
Can you please answer the following questions for me:
1) when they say it worked outside of WAS, did they write their own
program and tried to get a connection after the database went down? or
are they relying on some other mechanism to validate the failover.
2) will they eventually failover, or will WAS never failover.

Regards

0 new messages