My boss keep telling me that my company will use standby database. OK, here is the exactlly words- once our production database is down, backup DB will be picking up in 5 min after we apply logs. But he also say, two database (both production and backup machine) share same hard disk (they will share same datafiles..)- I am confused, is this standby? It doesn't look standby for me!
Can anyone give me some idea about that?
Oracle 126.96.36.199.1 (std) on NT 4.0
In a standby database, the dbf (from the backup) are copied to another server where the standby db will read the log files. So technically speaking, they will not share the same dbf. The only time the same dbf are shared are in the case of OPS.
You are right. If 2 instance share same datafile, once one datafile corrupted and 2 database corrupted as well.
The reason for them to design this is to prevent hardware failure only. Once datafiles coruptted, 2 DB server all need recovery. Than, from database perspective, where is our backup for db?
Could anyone give me some suggestion and I can tell my BOSS!!
Another option is the setup of Adv. Replication. There is a 3rd party tool called SharePlex by Quest which also reads log files onto a standby db.
[Edited by Halo on 05-01-2001 at 02:16 PM]
The configuration the Boss is apparently talking about is Parallel Server in the Oracle world, where two instances share the same dbf's/data files, but are running on different machines to safeguaud against cpu failure. Usually, mirrored data drives are involved. Depending on hardware, fail over can be as short as 1 minute (we have such a configuration). When CPU 1 goes down, there is software which switches the comm links to CPU 2, which trucks on down the road. The only loss are the transactions executing and uncommitted on CPU1. Since the instance on CPU 2 was up an running, there should not be any corruption issue, as the datafiles should not be housed on cpu 1, but in some kind of shared cluster.
The arrangement works.
The architecture predates what is now the 'standby' database approach, where a second system is set up and fed the archive logs from the first system.
Joseph R.P. Maloney, CSP,CDP,CCP
'The answer is 42'
Your Boss may also be referring to the cluster configuration where you have common disk and two nodes accessing it. All the disks will be mounted on the primary node and when that node crashes the secondary node will take over the disks.
As far as ORACLE is concerned, when the primary node crashes the instance will crash and when you access the disk from secondary node, you have to do an instance recovery. All these steps can be scripted and you can have smooth fail-over. The time to fail-over depends on the cluster software. You will not loose any commited transactions.
Hope this helps.
Click Here to Expand Forum to Full Width