Can fast start failover be achieved in maximum performance mode

Question

In a physical standby I would like to know if it is possible to configure High Availability in async mode to achieve fast-start failover? If so, please briefly describe how. Thanks

Answer 1

I've always said, size doesn't matter to Data Guard except when you are creating a standby database. After that it is the amount of Redo that is generated by the Primary database. A failover is a failover, for a 10GB database or for a 60TB database. The scary part is having to get the old Primary back up as a standby as quickly as possible.
Of course if the original Primary is completely toast then you have to recreate it with a new backup of the new primary or use a backup of the original Primary that happens to be on the Primary system and restore it as a standby and let Data Guard catch it up with the current state of affairs.
But, if the original Primary is still intact then you can use flashback database to reinstate it as a physical standby with little effort and no major backup restores. And of course, that is one of the requirements of Fast-Start Failover (FSFO). To enable it you have to have flashback database enabled on both the Primary and the FSFO target standby. Since enabling FB Database requires the database to be at the mount state I would suggest you enable it right after the upgrade to 11g.
When FSFO triggers a failover after a failure of the Primary, the Observer will automatically re-instantiate the failed Primary as a standby when it is started again and then it will be resynchronized with the new Primary automatically by Data Guard as normal.
The only difference between Maximum Availability FSFO and Maximum Performance FSFO is that you have to decide how much data you are will to lose when a failover occurs and tell the Broker by configuring the FastStartFailoverLagLimit property (see section 9.2.12 of the Broker manual).
You also want to be sure to place the Observer on a system where it is not affected by the Primary or Standby server going down or a network break between the Primary and the Standby. And if you can, use Grid Control. When you enable FSFO in Grid Control (on the Data Guard home page) you can specify 2 systems where the Observer can live. If it goes down on the original system an atempt will be made to restart it there. If that system if no longer available then Grid Control will attempt to start a new Observer on the 2nd system.
I would highly recommend you read the section [5.5 Fast-Start Failover|http://download.oracle.com/docs/cd/B28359_01/server.111/b28295/sofo.htm#i1027843] for lots of good information. And of course, the [Data Guard 11g Handbook|http://www.amazon.com/Oracle-Guard-Handbook-Osborne-ORACLE/dp/0071621113] (sorry, still plugging the book :^)
Larry