how to keep a disaster recovery db up to date

[John Flanagan]

I have set up a second box which will be kept in a different building for disaster recovery purposes.I will probably transfer archive logs over the network to do this. Does anybody do this and is it reliable. All help including scripts welcome. DB is Oracle 8.0.5.

Thanks,

John.

[Julio Yamawaki · Accepted solution]

Hi,

There's no secret on building a standby DB.
You simple have to make a full copy of the DB and apply all archivelogs.
In my case, I apply just one time a day, but you can make a crontab entry where you copy from the other box (using ftp in batch mode) and call a script to update the DB.
I don't have the script but if you need we can make an effort to develop this to help you.
Also, if you are a Metalink user, there are a lot of information about standy DBs in there.

[Ian Dennison_1]

John,

You are probably best to write some custom scripts. If you want to be tricky, you can get the scripts to check for assured delivery, sequence numbers, and failures within Oracle to apply redo logs.

We tried NFS but settled on gzip-ping the logs and ftping them.

The main issue with this is determining if something has gone wrong, and providing yourself (or the operators) the information needed. Sorry, can't provide my old scripts for this; previous employer has them.

Got a Storage Array in the mix? Can you use it's Replication facility (EMC = SRDF, HP = Business Copy)? Veritas does some nice incremental update software, maybe you can look at that?

Share and Enjoy! Ian

[John Flanagan]

Thanks for help. I will check metalink for more info.

Did you find this process reasonably reliable?

John.

[Julio Yamawaki]

Hi,

We are using this process to make our Disaster Recovery Plan.
We make a test every 6 months and it's completely secure.
In your case, if you make a update of archivelog, say, evey 30 minutes, in case of disaster, you have lost of data of 30 minutes (maximum).
Note that, you can make 15 minutes or less, but it depends on the speed where your archivelogs are generated.

[John Flanagan]

Thanks Yogeeraj, That is an excelet reply and contains much if the information I require.

Julio, do you abtain this reliability with a scripted solution or do import logs manualy?

Does my standby database also need to be configured in archivelog mode?

Regards,

John.

[Yogeeraj_1]

hi again,

The standby database can operate in NOARCHIVELOG mode. However, if ever you have to switch to it, you should immediate enable ARCHIVE LOG mode.

Again, remember that in 817 -- you can either deploy a managed standby where by the ARCH process delivers redo log files over the network to another system (requires enterprise edition) or you can move the archived redo log files yourself (non-managed standby).

In either case, the standby database is running in "recovery mode" -- meaning it is applying archives as they are recieved normally.

Also, note that standby database may be opened READ ONLY for queries and such in 8i. When the standby is opened read only, it is not applying archive redo log (eg: the logs are being collected but not applied).

As for "reliability with a scripted solution or do import logs manualy", i believe that it is up to you making it so. You have to ensure every thing yourself. Managed standby relieves you from this pain!

hope this helps!

Best regards
Yogeeraj

[Graham Cameron_1]

Before you start writing shedloads of scripts, Oracle have a product to do all this for you.
It is called DataGuard, is supplied with 9i but has been packported to 8i.
We use it on 8i on all our production databases - works fine.

- Graham

[John Flanagan]

Unfortunately we use 8.0.5. I think it is now time for me to work through all the documentation and get everyting up an running.

Thanks for the help everybody

[Carlo Henrico_1]

John

Just to add my 2 cents worth...

I make use of the extensively and in actual fact switch log files every 10 minutes. The Disaster Recovery database is never more than 10 minutes behind.

One major issue though is the following:

Regarding the format of the archive log files, please have a look at:

<60514.1> (This is on Oracle's Metalink) "ALERT: Archive Logs can be Silently Lost on Desktop Platforms"

I am using

log_archive_format = arch_%t_%s.arc

And it will continue to increase sequence number.

Good luck

Carlo

[Julio Yamawaki]

John,

Sorry, I was out of office today.
About our solution, it's partially made by scripts because we use this stand by db to disaster recovery and we make an update every night (this script simple get the archivelogs and update until the last - 1).
Whe have another script in crontab that copies archivelogs every 30 minutes (in case of disaster, this archivelogs are safe!).
Remember only to make logs in your scripts, to verify the successfull completion of the updates (this logs can be automatically send to your e-mail).

[Henrique Silva_3]

We are running 8.1.7, and I am implementing an Oracle standby solution for drp.

now, I wanted to also have a local standby server, close to the production one, where I could run my reports from, and do my backups from ( have not implemented RMAN yet ).

now,I was under the impression that I could only have one stand by db, but read here of multiple ones.

Do you have any docs where I can how to implement that ? I was going to use BCVs to get that local copy of production going, but if I can set up two stand bys ( one local, one remote ), then, this would be perfect.

Also, does anybody have any experience with omniback and rman integration ? I hear it is prety smooth !!!

cheers,

Henrique Silva

[John Flanagan]

Henrique Silva,

I can't answer these questions for sure but it sounds like you can have as many standby db's as you like if you use the archive log mode of update.

You should probably post a new question in the forum.

John.