HANA DB recovery options

[sumitjais]

Dear experts,

Looking for the feasibility wherein HANA database does not turn online upon the completion of database recovery, enabling to apply subsequent logs later.

Problem Statement:

Performing a system copy i.e. across data centers for a very large database, along with target HANA DB upgrade with backup-restore method.

Due to some limitations at source data-center for HSR , I am considering for the backup-restore approach, in which the full backup is restored much ahead the cut-over date with option -Recover the database to specific data backup. Until the time of cut-over, we wish to apply the subsequent logs,without having to restore with any data backup.

We can do such log recovery in databases such as MS SQL and Oracle as long as the database is not open (for connections), but HANA DB automatically comes online for usage after recovery.

[Foivos]

Sorry this may be a silly question, just need to understand your requirement better.

You want to do a system copy which is across data centres.

Your DB size is big (define big please, how many TB?)

What is the target system used for? Is this to become a QA system in your landscape, or is it for DR for example?

How long does the restore take if you do it including data?

[Foivos]

Sumit this is where I would start mate:

https://help.sap.com/docs/SAP_HANA_PLATFORM/6b94445c94ae495c83a19646e7c3fd56/843022db7b80427ea53b4e5...

You need to do a dress rehearsal of this to understand the timings needed though and it assumes you can connect across the Data Centres.

Take a look. Hope it helps 🙂

[sumitjais]

Thanks foivosg,

The database size is around 12TB. It's a DC migration of Productive system.

I am aware of tenant copy/move but i want to achieve this by backup/restore in a reduced business downtime.Hence, considering for the feasibility of applying subsequent logs i.e. after performing recovery with data backup few days prior to cutover.

[Foivos]

Sumit, if migration is what you are after you don't need a system copy mate. Just replicate across DCs. Near zero downtime... When you are ready to cutover, simply take over from the target DC and you are done. No system copy needed whatsoever.

Just FYI, I have done this with a 15TB HANA DB, across DCs and the replication from scratch took around 4.5 hours. Obviously this depends on the connection you have across DCs. You can start the replication whenever you want, it will keep the systems constantly in sync and when ready to cutover, you just takeover from the secondary site.

[sumitjais]

foivosg As mentioned earlier, there are limitations from source DC for setting up replication, hence looking for this approach.

[sumitjais]

lbreddemann denys.kempen Can i request for your expertise here please.

[Foivos]

Sumit, I don't think what you want to do is possible.

Please check Note: https://launchpad.support.sap.com/#/notes/1642148

Point 33 - Is online recovery possible?

"No. Database recovery is always a full recovery of the whole database and therefore always requires a downtime"

So you will not be able to restore, keep the DB on stand-by and replay the logs.

Can you explain what the limitation is in the existing DC? Do they not allow connections to other DCs?

[CharlesFeng]

Hi sumitjais,

As far as I know/checked, HANA DB will be online right after the recovery. "RECOVER DATA" command does not support such option.

Reference:
"After the recovery, the SAP HANA database is online and can be used by applications."

https://help.sap.com/docs/SAP_HANA_PLATFORM/4fe29514fd584807ac9f2a04f6754767/93637a07e3b544398aa02de...

Best regards,
Charles

[suryadeepnookala]

Hi sumitjais, Could you please let us know how did you come up with this issue? How did you replay the logs? Without an answer for such scenarios is nothing less than watching a horror movie without a climax. 🙂

[sumitjais]

Hi suryadeepnookala, thanks for the reminder 🙂

[sumitjais · Accepted solution]

Thanks all,

I explored the possibilities. Since the HANA DB does always come online (and is open for establishing connections) after recovery, we cannot apply further logs on it.

To apply further logs for roll forwarding, we must need a Full backup as base snapshot of the persistence.However, the subsequent recovery goes relatively faster as the recovery process just walks through the last save-points and roll forwards with the additional logs.