Re: RAC problem, is gc_files_to_locks is needed?
- From: "biti_rainy" <biti_rainy@xxxxxxxxx>
- To: "restomi_w@xxxxxxxxx" <restomi_w@xxxxxxxxx>
- Date: Thu, 21 Oct 2004 13:18:17 +0800
hi,tomi wijanto
from oracle document
Setting this parameter to any value other than the default will disable Cache
Fusion processing in Oracle9i Real Application Clusters.
from
Oracle Real Application Clusters
by Murali Vallath
Local
When the block is read for the first time into an and there are no other
instances in the cluster that have read the same block or are holding the
block, then the block is locally managed
Global
If the block that was originally acquired has been modified by the holding
instance and, based on a request from another instance, has transmitted the
block over. The block that was originally on one node is now present in
multiple nodes and therefore is considered globally managed
When an instance reads data blocks for the very first time, its existence is
local, that is, no other instance in the cluster has a copy of the same block.
The block in this state is called a current state block (XI). Therefore, the
behavior of this block in memory is similar to any single instance
configuration, with the exception that GCS keeps track of the block even in a
local mode. Multiple transactions within the instance have access to these data
blocks. Once another instance has requested for the same block, then the GCS
process will update the GRD, taking the state of the data block from a local
role to a global role.
Best regards
msn: biti_rainy@xxxxxxxxxxx
a dba from alibaba(china)
---- from the mail-----
>Hi All,
>
>I has oracle9i database with RAC using 2 nodes.
>The first time i used them, i activated server load
>balancing, so user connect to least loaded instance.
>
>But after i checked 'gcs/ges wait' related to cache
>fusion, that is very large, i activate only one node,
>and use another node as failover node.
>What i expected here was, i want to reduce
>intercluster operations for sql/dml.
>
>The problem i still got was, when one table was
>actively update by many users and i did full table
>scan on it, it's quite slow. I have compared it with
>non-RAC database, it's about 2-3 times slower.
>I have activated sql trace, and found many2 of 'global
>cr request' wait on that full table scan.
>
>So i feel a little confuse now. I guess when providing
>consistent view of updated blocks, the active node
>still check for existency of those blocks in another
>instance (that should be failed because the node is
>pasive), and then read from undo segments. That is
>probably why non-RAC database got faster.
>
>The question is: How could i reduce intercluster
>operation in full table scan. I have read about
>gc_files_to_locks parameter, but i'm afraid of
>inflexibility caused by it (that i need to bounce
>server to change it).
>Is there any best practice to use gc_files_to_locks?
>I'm really curious on this parameter and wondering if
>it's worthed to considered..
>
>Best regards,
>tomi
>
>
>
>
>__________________________________________________
>Do You Yahoo!?
>Tired of spam? Yahoo! Mail has the best spam protection around
>http://mail.yahoo.com
>--
>//www.freelists.org/webpage/oracle-l
>
>.
--
//www.freelists.org/webpage/oracle-l
Other related posts:
