Re: TAF with Oracle RAC
Date: Fri, 24 Jul 2015 11:07:17 +0200 (CEST)
Message-ID: <2016280103.6915.1437728837215.JavaMail.open-xchange_at_app02.ox.hosteurope.de>
Hi Chris,
just for clarification. You described two different scenarios, so i am a little bit confused.
Scenario A) "I then opened a session to RAC and stopped the instance which the session was connected to, but there was no change in behaviour. I got an ORA-03113 at the client connection when I reissued the select statement, and then reissuing the select again gave me normal output."
Scenario B) "The test results now show that the application log received an error, but the user interface did not display any errors and it continued processing the long query thru to completion, even though I'd brought one node down."
Scenario A is called "session failover" and scenario B is called "select failover". I am just qouting from Martin Bach's book "Pro Oracle Database 11g RAC on Linux" ( http://www.apress.com/9781430229582 ) again as it is one of the best resources for such stuff:
- The session failover mode is probably the most basic case because it doesn’t require any code changes for implementation. All that is required is a suitable local naming definition in the form of a tnsnames.ora file or a TAF policy attached to a database service. TAF will automatically reconnect the user session using the same connection string if a node failure occurs.
- The select failover mode allows clients to continue fetching rows from an open cursor after the session has been re-established on another node. Internally, TAF will re-execute the same query, discarding the rows already returned to the user and thereby making the failover transparent. There is a slight overhead associated with this because Oracle has to keep track of which data has already been transmitted.
So i am not quite sure what you wanted to test with scenario A, but it seems like you focused on two different scenarios. By the way my 12c demo was about scenario A.
Best Regards
Stefan Koehler
Freelance Oracle performance consultant and researcher
Homepage: http://www.soocs.de
Twitter: _at_OracleSK
> Chris King <ckaj111_at_yahoo.ca> hat am 23. Juli 2015 um 20:29 geschrieben:
>
> Thanks everyone for all your input on this issue. I think it's resolved now.
>
> I've added the TAF service on the server side as per Svetoslav's recommendation and changed the method to BASIC. The test results now show that the
> application log received an error, but the user interface did not display any errors and it continued processing the long query thru to completion,
> even though I'd brought one node down. That's what I was looking for.
>
> Many thanks!
> -ck
-- http://www.freelists.org/webpage/oracle-lReceived on Fri Jul 24 2015 - 11:07:17 CEST