It is crucial to monitor wait events, global enqueues, and system and session statistics as significant waits inevitably indicate a performance problem. If a wait is in the top five list of an AWR report, it needs further investigation by drilling down to the event, that is, the reason of the wait, while bearing in mind that these are global events and all the instances of the RAC cluster are involved.
We need to avoid so called "one node myopia" syndrome and troubleshoot the problem from the concept of RAC as a cluster database with all the instances of the cluster. ADDM, AWR, and ASH reports are instrumental, and most of the time sufficient to start diagnosing and troubleshooting performance problems in RAC. Performance V$ views are also very useful in providing thorough information for troubleshooting and performance tuning.
Each RAC instance takes its own AWR snapshot at approximately the same time. Each instance...