Query on Duplication and Replication

[Dav1234]

Hello Experts,

Here, in my infra ..

The first backup copy goes to appliance and then it is duplicated to another Netbackup appliance and post that 3rd copy is replicated for LTR.

what i am see here is backup failed(first copy) but it is still duplicating and replicating..Is this normal process??

Because i think, if the first copy is not completed then the 2nd copy(duplication) and 3rd copy(Replication) is niot started.

Please confirm. 

[davidmoline]

Hi @Dav1234 

Secondary operations can only start after a successful first operation (i.e. the backup). I suspect what you may be seeing is a SLP backlog or if the "backup" comprises many parts - some part may have succeeded and so will continue to be processed by the SLP. 

By a backup of many parts what I am referring to is a multi-streamed backup where only some streams fail which then the overall backup "fails" even though there have beeen successful streams. 

SLP operate at the backup image level. 

Hope that explains things
David

[Dav1234]

Hello David,

Understood, lets take  example of RMAN backup wherein, 3 streams got failed and rest 8 streams completed successfully but this backup is of no use as i cant recover the database from this partial backup(as 3 streams got failed and  parent stream failed)

so, is there any mechanism where 2nd copy(duplication) and 3rd copy(Replication) will only start if backup completed successfully(including parent stream).

[davidmoline]

Hi @Dav1234 

I'm not aware of any mechanism to achieve this (it would be nice I agree). 

About the only way you could look to do this would be to delay the secondary operations until you have confirmed a successful backup. If the backup fails, you could then expire (force) the bits that did succeed (but are otherwise not very useful). It potentially could be scripted with some thought. 

The SLP delay could be achieved by SLP windows, or increase the max size for SLP duplications or increase the time for force interval for small jobs, or even creating an SLP for the job that by default has secondary operations suspended.

Cheers
David

[Dav1234]

Hello David,

Thank you for 2nd option but more manual work...

Would it be OK to open a case with Vendor to check for alternate, which we disussed?

And if such a mechanism exists, will it not have any impact on the performance of duplication and replication?

[davidmoline]

Hi @Dav1234 

Of course you may open a case - but I don't believe they will be able to provide any help beyond what I have suggested. Then again I may be surprised (or someone els eon this list may have alternate suggestions). 

Cheers
David

[Michal_Mikulik1]

Hello,

I am sure there is no NetBackup supported solution for this. Keep it as it is, because you are not 100% right that partial RMAN backup is useless:

- you can use partial backups for datafile-level loss

- you can use it also for the whole database restore, for example rman is able to use 20 datafiles from older backup and remaining from newer backup to recover the whole db

- RMAN is assessing success of full backups and incremental backups on datafile level, so with not copying partial backups to next SLP copy, you could break full/incremental chain there. Similar idea like when you would expire primary backups for partial backups presuming that they are of no use

Regards

M.

[Dav1234]

Hello Michal,

Means we can recover whole database with partial RMAN backup?

[Michal_Mikulik1]

Hello,

if you have at least 1 backup of each datafile, regardless the age of backup, and all archivelogs covering the whole backup history timespan, then yes. It is a good point to employ also incremental backups to speed up the recovery.

Example: you have partial backup from Mar 10th, and full backup from Mar 3rd. You have all archivelogs covering Mar 3rd-Mar 10th. Then yes you can recover the db to the state from Mar10th.

If you will have also incrementals from every day for example, you can speed up the restore process substantially, but this is not a must.

regards

M.