Recovr point -oracle crash consistent image

See EMC® RecoverPoint Replicating Oracle with RecoverPoint Technical Notes available on Powerlink.

In short, all RecoverPoint snapshots are crash-consistent, as long as you put data files, control files, and online redo logs into the same RecoverPoint Consistency Group.

we have all of them in same CG but when we try to  recover from any  of the snapshots (not the hotback one) it give fuzzy error in data base

Since you can recover from hot backup, it seems the problem is on the Oracle side.  Oracle should be able to come up if it lost power, connection to storage, crashed, etc.  Other than going over the Tech Note, I am not sure what could generate a fuzzy error.

when we try to recover the database from a snapshot which is not  part of hotbackup. while recovring the db it give fuzzy errror, so the problem is lyki this

we run the hotbackup script twice a day and create the bookmark. and we can only recover our db from these two bookmark. suppose in case of the disaster if the last bookmark was created at 2pm and disaster occurs at 8pm .

so i can only recover from 2pm image as this only consistent image i have. so i loose the data after 2pm

even though i have latest snapshot of before 7:59 PM.

so its like a point in time copy so why we do say it continous data protection. when i can recover it from a point in time because other point are not relvant

I am sorry if I was not clear in my previous post.  RecoverPoint provides continuous crash-consistent points in time and Oracle can and does recover from any of those points in time.  The fact that it works with hot backups and does not work otherwise tells me that the problem in your specific environment is on Oracle side.  As documented:

Oracle has developed reliable, largely automatic procedures for recovering its database files following a hardware crash, instance crash, or process crash. For a full description of these procedures, refer to the Oracle Backup & Recovery Handbook for the version you are using.

Oracle crash recovery regularly involves the following components:

• Data files

• Online redo logs

• Database control file

If a database is shut down normally, the process of getting to a point of consistency during restart requires minimal work. If the database terminates abnormally, restarting will take longer, depending on the number and size of transactions that were in flight at the time of termination. A RecoverPoint disaster recovery image without Oracle hot backup is similar to an abnormally terminated database. The Oracle restart procedure transforms the replica image into a transactionally consistent image by completing committed transactions and rolling back uncommitted transactions during database initialization.

In recovering a database after an instance crash, Oracle uses this data in the following two-step process:

1. Roll forward

Oracle sequentially applies redo records, as necessary, from its online redo logs to the corresponding data blocks. This ensures that the recovered database contains all committed data written by the application prior to failure of the database.

2. Roll back

Oracle chains through the undo data for each uncommitted transaction and rolls back all changes to the database accordingly. This process, also called transaction recovery, ensures that the recovered database is consistent on the transaction level; for example, that it does not contain a debit without a corresponding credit.

Though this process does not by itself constitute a disaster recovery solution, the same process enables seamless recovery when a RecoverPoint point-in-time snapshot is used.

[. . .]

I am not sure which procedure you are using to restart the database and which error you are getting.  I believe all you need to do it "startup", but you can get much more detail in the Technical Note.  The bottom line is that it should work and it does work at many customer sites.

"At the DR side we have taken clones from the RP replica."  Are you using Logged / Physical Access before cloning the replica?

"when i try to refresh it from non hot backup bookmarks ie normal snapshots it give fuzzy error."  I am not sure what the error is.

"If the disaster occur and i have to bring up the environment first at DR site can the DB be recovered from the normal snap shots or just from the hotbackup bookmarks. [. . .] when we can recover the db from any snapshots then what is the use of putting the DB in hotbackup mode and creating the bookmark."  It is better, more graceful, to recover from a hot backup snapshot, but you should be able to recover from any snapshot.

Hi Dmitry,

We tested the crash consistent recovery at our environment, it was a big success.

We were able to recover the Data base from the latest image (non hot backup Image).

And the DB recovered  without using archives.

Thanks a lot.

But can you explain when we should we use the hot backup and when should we use

crash consistent .

-Shailendra

Shailendra,

With unlimited crash-consistent snapshots you have a piece of mind that you can recover your database at any point in time.  This is very useful, if you have an unexpected problem, such as data corruption caused by a unintentional user error, problem in applicaiton logic or even a security breach.  RecoverPoint allows you to go back in time rigfht before the problem occured.

If, on the other hand, you have a planned activity, such as an online upgrade, you will want to generate a bookmark (i.e. name a snapshot), right before the activity.  Since you know when you plan to start the upgrade, it makes sense to create an applicaiton consistent bookmarked snapshot by briefly putting Oracle into a hot back mode.

Finally, it's a good practice to generate an application consistent snapshot once a day, just in case.

Hope it helps,

Dmitry

This is something I am currently looking at - how to provide transactional consistent copies of multiple Oracle DB on RP.

Putting the DB into hot backup mode on a frequent basis - lets say my RPO is 10 mins - is not an option.

So we are about to start testing using the crash consistent bookmarks to recover the copy (CRR)

In your opinions has this ever been a cause of concern wrt losing data?

Hey,

We have been using  Crash Consistent solution for past two month after Dmirty Suggestion.

It worked for all Databases including OLTP and OAPS.

We were able to recover from 2 min back image in CRR. Its a great success.

Though we are still creating Bookmark images as its suggested by EMC.

Crash Consisten image work in all cases except in the case when mount point are changed .

In that case we need to go with hot backup image and DB is recovered by applying archives.