Error message with Oralce 9.0.1.4 RAC with 2 instance

[ericfjchen]

This is Oracle 9.0.1.4 RAC with 2 instances.
While instance 2 abnormally down, the alert files are as below,
-----------------------------
instance 1 alert.log....
Wed Jul 20 18:28:27 2005
Reconfiguration complete
Wed Jul 20 18:28:27 2005
Instance recovery: looking for dead threads
Instance recovery: lock domain invalid but no dead threads
---------------------------
instance 2 alert.log....
Wed Jul 20 18:28:27 2005
Reconfiguration complete
LCK0 started with pid=17
Wed Jul 20 18:28:33 2005
Successful mount of redo thread 2, with mount id 912391497.
Wed Jul 20 18:28:33 2005
Database mounted in Shared Mode (CLUSTER_DATABASE=TRUE).
Completed: ALTER DATABASE MOUNT
Wed Jul 20 18:28:33 2005
ALTER DATABASE OPEN
--------------------------------
Why did instance 1 show ....
Instance recovery: looking for dead threads
Instance recovery: lock domain invalid but no dead threads
What's going on?

Thanks

Eric

[Joseph Loo]

hi,

may i know what is db_block_size? also, is this the first time it encounter the problem?

have u also apply any patch with regard to this problem?

u may also like to do something about your assigment of points, it is 3 pages and counting:

http://forums2.itrc.hp.com/service/forums/pageList.do?userId=CA951816&listType=unassigned&forumId=1

regards.

[Indira Aramandla · Accepted solution]

Hi Eric,

Refer MetaLink Doc ID: 188704.1

This message is just for information. If you don't use replication, you simply can ignore it.
The "replication_dependency_tracking" message is normal acc to Metalink document. "'REPLICATION_DEPENDENCY_TRACKING turned off' message seen in alert logfile" for details on this message.

The parameter REPLICATION_DEPENDENCY_TRACKING enables or disables dependency
tracking for read/write operations to the database. Dependency tracking is essential for propagating changes in a replicated environment.

The parameter REPLICATION_DEPENDENCY_TRACKING which is probable set to TRUE (default setting) inside INIT.ORA or SPFILE. This parameter has been introduced with Oracle8i, however no message was logged in alert logfile with this release.

Lock Domain is a catch all term for resources and DLM locks. This information is distributed among all the active nodes. When instances start or are stopped, This information needs to be redistributed across the new set of active instances. This is done automatically and there is no need for manual intervention.

This is a very common message and can be ignored unless followed by any 'ORA-xxx' errors.


Indira A

[Ariel Cary]

Eric,

From the alert logs you showed, it seems you tried starting up the inst2, the one that terminated abnormally, at that point the instance recovery was carried out by inst1, the surviving instance. Essentially, the SMON process of the surviving instance remasters all locks held by the departing-joining instance. In this case this was executed by inst1's SMON process, and that's what you see in the alert log of inst1. This is expected behavior and is a difference in RAC environments.

HTH,

ARC