I have two Oracle 11g Database servers, one at primary site and second at failover site, and i intend to use Oracle Data Gaurd for failover in between. I have few questions 1) Do i need standby database back up. If so, would the requirements be same as of primary? 2) In case of failover, what is good, automatic failover or manual? 3) What configuration and implementation setup is required to maintain a common IP address (Virtual IP) to the databases.
despite much admiring Chen's excellent blog, I have to say I disagree on question 1 above.
Backing up your dbs via RMAN should just be standard practice. In this case, you could argue the standby system is your db backup, which is fine. But it does expose the system to more risk - for example, if the standby is unavailable for whatever reason, and you're not taking RMAN backups on the primary. Moreover backing up both dbs allows you to prove that both sites work OK, which again lowers risk.
Unless there's a clear and compelling reason not to, it's safer to just back up both sides (and do periodic switchovers to prove your operational documentation is still valid).
For point 2) this partly depends on your uptime requirements - can the system afford to wait for someone to run manual commands for instance? There is no right or wrong answer on this but for our systems we do run things manually, which gives us a chance to check things before taking action. And it lessens the risk that something like a network blip could trigger a failover unnecessarily.
On point 3) you can indeed use a DNS alias, or you have the option of having the new primary db register a service with the listener. Both work fine, the latter only needs DBA input to set up.
Cheers,
Mark
1) No. You should backup either primary or standby but not both. It is recommended to backup standby only to reduce load on primary. Since standby and primary have the same DBID you can use the backup to restore either.
2) By automatic do you mean FSFO or using dataguard broker? Using the broker is great - it is one command and it checks that everything is working before doing the failover. Much easier and safer than manual. FSFO is complicated and has huge administrative overhead so only use that if you really need to.
3) We don't use virtual IPs. We have a DNS name for the DB. The app servers and clients use the DNS name. In case of emergency, we modify the DNS and flush the configuration on the app servers and clients.