RAC: Global Cache Converts & V$SYSSTAT "Time"
- From: "Marquez, Chris" <cmarquez@xxxxxxxxxxxxxxxx>
- To: <oracle-l@xxxxxxxxxxxxx>
- Date: Thu, 20 Oct 2005 17:43:19 -0400
9205 RHEL3ES RAC Are these numbers as bad as they look. I want make sure I'm not over reading or incorrectly computing here. Also little confused out the true meaning of "time". ---8i, 9i & 10g doc say; global cache convert time Real Application Clusters only: Total time elapsed during lock converts Would nice if the define time!? col "Average Time" format 999.999 SELECT V$SYSSTAT.VALUE*10 / (select value from v$sysstat where name = 'global cache gets') "Average Time" FROM V$SYSSTAT WHERE V$SYSSTAT.NAME = 'global cache get time' / Average Time ------------ 19.883 col "Average Time" format 999.999 SELECT V$SYSSTAT.VALUE*10 / (select value from v$sysstat where name = 'global cache converts') "Average Time" FROM V$SYSSTAT WHERE V$SYSSTAT.NAME = 'global cache convert time' / Average Time ------------ 390.042 I believe I have a real interconnect/RAC performance issue here. The sad part is that one node is idle...no end users (oracle) connections at all. Some jobs maybe run on this node and the RMAN backups, but that is about it!...meaning very few (reasons) to "get blocks" from the other node at all!? We had the idle failover RAC node down for a few days and I brought it back up last night. Also, I there are almost no "global cache" related waits (today) on the active node...yet all day the load on the active node it high. Ultimately we suspect our interconnect but I don't know how to go about verify its performance? PS Yes we have implement the recommend udp send / receive OS parameters. Thanks, Chris Marquez Oracle DBA
Other related posts:
- » RAC: Global Cache Converts & V$SYSSTAT "Time"