RAC – It’s all about the node affinity
January 16, 2013 6 Comments
Or what happens when you don’t have node affinity.
Here’s an old, short post that was lying around unpublished.
SQL executed on the 14th on node 2, on 15th on node 1.
|DT||SQL ID||Node||Rows||Execs||Elapsed||Cluster Wait Time|
Timings in seconds.
32625 seconds is just over 9 hours, of which 32175 is cluster wait time.
Database services and application datasource config have now been reconfigured to the extent that there is node affinity and the data required for the workload on node 1 is completely segregated from the workload on node 2.
If you use RAC as a black box and you have problems with performance or problems with predictability of performance then look into whether node affinity, or a lack thereof, is your problem.