|
|
|
|
|
|
|
|
|
Re: log file sync and LGWR wait on LNS Event - ADDM [message #502793 is a reply to message #502785] |
Sat, 09 April 2011 13:55 |
John Watson
Messages: 8960 Registered: January 2010 Location: Global Village
|
Senior Member |
|
|
I don't think ADDM was any use for RAC back with 10g.
Those global cache buffer busy and current block busy wait events are the ones I would look at. They are giving you an awful CPU time to DB time ratio. Looking at the "SQL ordered by cluster wait time" section you can see the code that is hitting the problem: those insert statements, which are giving you the block contention between the instances. I would look at the segments involved. Can you reverse key the indexes? Or hash partition the tables? And of course, are you using automatic segment space management?
|
|
|
|
|
|
|
|
|
Re: log file sync and LGWR wait on LNS Event - ADDM [message #502837 is a reply to message #502826] |
Sun, 10 April 2011 14:32 |
John Watson
Messages: 8960 Registered: January 2010 Location: Global Village
|
Senior Member |
|
|
Hash partitioning your tables and indexes might reduce the contention, if concurrent inserts will then go into different partitions. You'll need to find a partitioning algorithm that will result in some partitions having an affinity to each instance. But you have to be absolutely certain of what will happen. A partitioning strategy that works for your inserts might be a disaster for your queries.
As for memory, well, I always do what the advisors tell me to do.
|
|
|