| Oracle FAQ | Your Portal to the Oracle Knowledge Grid | |
 |
 | |||
| Oracle FAQ | Your Portal to the Oracle Knowledge Grid | |
|
| |||
Home -> Community -> Mailing Lists -> Oracle-L -> Re: DataGuard
David,
In my opinion DG is a nice solution for HA/DR. What I particularly like is
the possibility to have your transactions transmitted synchronously to the
standby, while applying the redo can be postponed for several hours. This
protects you from human errors: when someone drops a table/deletes all
rows/fill in your error with some human cause, you can recover the standby
up to just before the error, open it in read-only mode and recover the data
from the standby. Aftwerwards, the standby continues as before. You can
also use th standby in read-only mode, and still have your redo shipped to it.
This is something that clusterfilesystems, replication, mirroring, timefinder and all the fantastic storage-based replication methods do not have. Apart from that, hardware mirroring takes far more bandwith. The minimum replication unit is a block, if not a complete track, depending on the HW involved. Replication will occur for online redolog (redundant redo counts twice), archiving and, of course, the datafile writes. DG sends just the redo data to the other end.
One of my customers has two machinerooms in the same business area, some
300 meters separated, connected through fibre. We have configured
transparent application failover, and a controlled switchover takes just a
few minutes.
James Morle is implementing an N+1 cluster. Several nodes are running
databases and applications. If one fails, the standby node will instantly
take over, including (virtualized) IP-addresses, SAN connections etc.
Here in the Netherlands a bank implemented a setup where one server runs the database, but another node has all disks mounted as well, but sits idle. Some home-grown scripts monitor the database server, and when it fails, it will start the instance and take over. I call this PMR: Poor Man's RAC. Saves you from the 50% extra Oracle license, and, when the standby node doesn't run for more than 10 days in a calender year, you don't need licenses for it at all.
I've setup over 20 DG configurations now 9 sites, in all modes. I think it is quite stable, at least Physical Standby is. Logical Standby in 9i isn't suitable for DR/HA. In 10g it gets a new chance, but it hasn't proven itself yet.
Ease of installation: I've build up my own configuration toolkit which I use in my contracting work. I was forced to build this because the GUI-interfacing didn't work in the first release, and the linemode data guard tools didn't have all necessary features. After some first configuration work in init.oras and Oracle Net instantiation is just a one-liner at the shell-prompt. In my opinion, the technology of Data Guard is quite straight forward and not too complex.
Regards, Carel-Jan
===
If you think education is expensive, try ignorance. (Derek Bok)
===
At 12:01 AM 7/8/2004, you wrote:
>I am looking for feedback on Oracle Data Guard in terms of stability and
>ease of intallation, setup and administration. I have some light
>experience on the product, but no real world intense experience.
>
>I am involved in a rollout of large RAC systems, but the drive force
>behind going with RAC if for fall-over fault tolerance, not scalability or
>performance.
>
>The questions is now being raised on if that is the best decision and what
>our other options are(Veritas cluster manager, Data Guard, etc).
>
>Thanks!
>- David
-- Archives are at http://www.freelists.org/archives/oracle-l/ FAQ is at http://www.freelists.org/help/fom-serve/cache/1.html -----------------------------------------------------------------Received on Wed Jul 07 2004 - 18:06:01 CDT
 |
 |