Martin Bach did a good study of Rac One Node capability and especially of what happens with session failover during a database relocation or during a node failure into this blog post.
He showed us that:
- during a database relocation (initiated manually with srvctl relocate database) the TAF works as expected.
- during a node failure the TAF did not work and you will be disconnected (for a running select or a new execution)
It’s important: I said “during a node failure”, it means no instances are available during the select.
But what’s about a session already connected to a TAF service before the node failure that :
- will be inactive during the node failure
- will launch a select after the node failure (means an instance is back)
For this one an node failure is transparent, let’s see that in action:
First check my database is a Rac One Node:
srvctl config database -d BDTO | grep -i type
Type: RACOneNode
Now let’s check that the bdto_taf service I’ll connect on is a TAF one:
srvctl config service -s BDTO_TAF -d BDTO | egrep -i "taf|failover"
Service name: BDTO_TAF
Failover type: SELECT
Failover method: BASIC
TAF failover retries: 0
TAF failover delay: 0
TAF policy specification: BASIC
Well everything is in place to test.
So connect to my Rac One Node database using the TAF service:
sqlplus bdt/"test"@bdto_taf
SQL*Plus: Release 11.2.0.3.0 Production on Thu May 16 14:39:36 2013
Copyright (c) 1982, 2011, Oracle. All rights reserved.
Connected to:
Oracle Database 11g Enterprise Edition Release 11.2.0.3.0 - 64bit Production
With the Partitioning, Real Application Clusters, Automatic Storage Management, OLAP,
Data Mining and Real Application Testing options
SQL> alter session set nls_date_format='YYYY/MM/DD HH24:MI:SS';
Session altered.
SQL> select sysdate||' '||host_name from v$instance;
SYSDATE||''||HOST_NAME
--------------------------------------------------------------------------------
2013/05/16 14:39:47 BDTSRV1
Now let’s kill smon:
ps -ef | grep -i smon | grep -i BDTO_1
oracle 8035 1 0 14:30 ? 00:00:00 ora_smon_BDTO_1
kill -9 `ps -ef | grep -i smon | grep -i BDTO_1 | awk '{print $2}'`
Let’s wait an instance is back on a node (Important step)
In this case the instance has been relocated on the other node, let’s see it is up:
ps -ef | grep -i smon | grep -i BDTO_1
oracle 29266 1 0 14:54 ? 00:00:00 ora_smon_BDTO_1
and come back to our “idle” sqlplus session to re-launch the sql:
SQL> /
SYSDATE||''||HOST_NAME
--------------------------------------------------------------------------------
16-MAY-13 BDTSRV2
Oh ! As you can see the “idle” sqlplus session has not been disconnected and we have been able to re-launch the query.
As you can see we lost our nls_date_format setting. This is expected as explained into Martin’s Book “Pro Oracle Database 11g RAC on Linux”:
”
Also, note that while TAF can reauthorize a session, it does not restore the session to its previous
state. Therefore, following an instance failure, you will need to restore any session variables, PL/SQL
package variables, and instances of user-defined types. TAF will not restore these values automatically,
so you will need to extend your applications to restore them manually.
”
Conclusion:
If you are using a TAF service to connect to a Rac One Node database (of course through the OCI), then a node failure is transparent if the session is “idle” during the time there is no instance available.
When you think of it, it is quite normal as the TAF is more or less nothing more than a OCI / SQL*net feature, so once the TAF service is back, then TAF can work as expected.
Remarks:
- If the “idle” connection has an “opened” transaction, you’ll receive the “ORA-25402: transaction must roll back” oracle error (This is not so transparent anymore :-) ).
- Do not test TAF connected as the oracle sys user as SYSDBA Sessions do not failover with SRVCTL TAF configured, see MOS [ID 1342992.1]
- If you are using srvctl to test TAF’’s reaction on node failure then you should read the MOS note [ID 1324574.1] first (as the TAF may not work depending of the version and srvctl options being used)
- For a good explanation of Rac One Node, you should have a look to this Aman Sharma’s blog post.
Why I wrote this post ?
Because into Martin’s post, I put a comment on ”December 13, 2012 at 19:08” trying to explain this behavior. I came back to this comment yesterday and I realized that my comment was not so clear ! I hope this post is clear ;-)