Backup an MAXDB with snapshot technology

[Former Member]

I installed an ContentServer on Linux64 bit. (MAXDB 7.6.05 B12)

The directories are installed on a netapp. Unfortunaly I haven't experience with netapp.

Now I hope to get some informations about backup and restore with snapshot.

Till yet I only have to do backup und recover with networker tools. What is to do if I have to restore

the content server to a specific time? (snapshots of data and log )

Best regards

Hermann Duessel

[Former Member]

Hello to all,

thanks a lot for your answers.

I thinks we have enough answers to go about

this topic.

[Former Member]

Hello Mr. Duessel,

yes, this issue should be discussed in an OSS call. Please open a call on BC-DB-SDB to get an official statement and recommendation by SAP.

Regards, Christiane Hienger

[Melanie]

Hello!

You might find note 912905 helpful, which points to this document:

http://media.netapp.com/documents/tr-3689.pdf

If you are searching for notes on component BC-DB-SDB* with keyword snapshot, you'll find further helpful notes (e.g. 371247).

Best regards,

Melanie

[Former Member]

Hello Mr. Duessel,

you will find detailed information about Snapshots created by the MaxDb database in the MaxDb documentation using the following link:

[Snapshot|http://maxdb.sap.com/doc/7_6/42/0e1a62fc297430e10000000a114cbd/frameset.htm]

Hope this will answer your questions.

Regards, Christiane Hienger

[Former Member]

Hello and thanks a lot for your answer by return.

But this topic I believe isn't the solution I mean.

At the time we bachup the Contentserver once a day ba save data and several logbackups using

legato networker.

To get lower costs we want to save the data and logs in the future not with legato networker by saving into a pipe and tape, instead of this our backup-team used the snopshot-technology on a netapp fileserver. If we backup an oracle-database we get an consistence of the data by setting the tablespaces to begin backup and after the snapshot to end backup.

The Structure of our contentserver is the following :

drwxrwxr-x 12 sdb sdba 4096 2009-01-13 09:42 .

drwxr-xr-x 29 root root 4096 2008-10-06 10:36 ..

drwxr-xr-x 6 sdb sdba 4096 2009-01-13 10:19 CSC

drwxrwxr-x 11 sdb sdba 4096 2009-01-13 09:43 data

dr-xr-xr-x 14 sdb sdba 4096 2009-01-13 09:43 programs

drwxrwxrwx 5 root root 4096 2009-01-20 01:00 .snapshot

drwxr-xr-x 12 sdb sdba 4096 2009-01-13 09:42 TS0

/sapdb/TS0>

dr-xr-xr-x 11 sdb sdba 4096 2009-01-13 09:43 db

drwxr-xr-x 3 sdb sdba 4096 2009-01-19 15:18 sapdata1

drwxr-xr-x 3 sdb sdba 4096 2009-01-13 09:43 saplog1

drwxrwxrwx 3 sdb sdba 4096 2009-01-20 10:48 sapwork

drwxrwxrwx 5 root root 4096 2009-01-20 01:00 .snapshot

/sapdb/TS0/sapdata1

drwxr-xr-x 12 sdb sdba 4096 2009-01-13 09:42 ..

-rw-rw---- 1 sdb sdba 20971520000 2009-01-20 10:48 DISKD0001

-rw-rw---- 1 sdb sdba 20971520000 2009-01-20 10:19 DISKD0002

-rw-rw---- 1 sdb sdba 20971520000 2009-01-20 10:19 DISKD0003

-rw-rw---- 1 sdb sdba 20971520000 2009-01-20 10:19 DISKD0004

drwxrwxrwx 3 root root 4096 2008-10-07 17:42 .snapshot

/sapdb/TS0/saplog1

-rw-rw---- 1 sdb sdba 4194304000 2009-01-20 10:43 DISKL001

drwxrwxrwx 2 root root 4096 2008-09-19 14:45 .snapshot

Now our backupteam will execute daily a snapshot of /sapdb/TS0/sapdata1 and several times

a snapshot of /sapdb/TS0/saplog1.

The expire date is 3 Weeks.

In a normal situation (saving Data by networker) I will execute a restore like this :

Restore Data and open the Database in case of save data is enough for the customer. Either I took an Online Backup (Complete Date) the backup is consitent.

Restore Data + restore several logbackups if the customer wish an just in time restore. Also in this case we have consitent system because the restore only takes the commited transaction.

How can I get the same result with the snapshot technology written below ?

Now our backupteam will execute daily a snapshot of /sapdb/TS0/sapdata1 and several times

a snapshot of /sapdb/TS0/saplog1.

In this case all transactions are frozen (commited , rollback ...)

I can also open an OSS-Call if you think that would be better .

Thanks for your help.

Best Regards

Hermann Duessel

[lbreddemann]

To get lower costs we want to save the data and logs in the future not with legato networker by saving into a pipe and tape, instead of this our backup-team used the snopshot-technology on a netapp fileserver. If we backup an oracle-database we get an consistence of the data by setting the tablespaces to begin backup and after the snapshot to end backup.

Just to set this straight: by using Oracle BEGIN BACKUP / END BACKUP you never get consistent backups. By defintion.

Without the archivelogs you cannot even open a restored database from these backups.

The BEGIN/END BACKUP" command is just there to let Oracle know that the files are going to be inconsistent and to change the logging behavior accordingly.

Now our backupteam will execute daily a snapshot of /sapdb/TS0/sapdata1 and several times

a snapshot of /sapdb/TS0/saplog1.

??? Snapshots of the logarea are as useless as snapshots of Oracles online redo logs.

Sorry - the only way to save your log entries is: perform log backup.

Oracle does the same - it archives the logs.

The expire date is 3 Weeks.

In a normal situation (saving Data by networker) I will execute a restore like this :

Restore Data and open the Database in case of save data is enough for the customer. Either I took an Online Backup (Complete Date) the backup is consitent.Restore Data + restore several logbackups if the customer wish an just in time restore. Also in this case we have consitent system because the restore only takes the commited transaction.

By performing snapshots of the log area you don't have any log backups yet - so there is no recovery possible with your solution...

How can I get the same result with the snapshot technology written below ?

MaxDB supports the a suspension of the I/O that makes it possible to have a I/O consistent state on the disks for the time required to create the snapshot on the storage layer.

Check notes

715666 Suspend/Resume Logwriter

616814 Suspend Log Writer für Split Mirror bzw. Snapshot

for that.

In this case all transactions are frozen (commited , rollback ...)

Of cause they are. But just for the short time the storage system needs to perform the snapshot.

Usually that's only a few seconds, and content server databases don't have a high transaction load profile.

I can also open an OSS-Call if you think that would be better .

In contrary to my colleagues I don't see the bug here - so I think the SDN is the right place for this discussion.

BTW: please keep in mind that a snapshot is by far not the same as a backup. You don't get the check for corrupt pages, you cannot easily perform a system copy to a different instance, you've far less monitoring etc...

Keeping up the same quality of backup solution will require quite a lot additional work here.

regards,

Lars