Networker

Re: [Networker] Basic questions about backing up Oracle databases

2011-09-09 14:27:58
Subject: Re: [Networker] Basic questions about backing up Oracle databases
From: "Clark, Patti" <clarkp AT OSTI DOT GOV>
To: NETWORKER AT LISTSERV.TEMPLE DOT EDU
Date: Fri, 9 Sep 2011 14:27:03 -0400
We've moved completely away from any of these backup devices and have gone 
completely to using LVM snapshots very successfully.  The "outage" for the 
database is as long as it takes to perform a snapshot - putting the database 
into "backup mode" for that part of the process and then returning the database 
to full operation.  It takes seconds.  Then using your favorite process for 
managing the database files, gzip in our case, put the files into a location 
for going to tape which can occur at any time.  Once the archive is completed, 
the snapshot is released.  The slight performance hit while the snapshot is 
present is much less than the one the database otherwise encounters under any 
other method.  We've done this for Oracle and now we are doing it for 
PostgreSQL.  Recovery is fairly straight forward, too.

Patti Clark
IIa

 

>-----Original Message-----
>From: EMC NetWorker discussion 
>[mailto:NETWORKER AT LISTSERV.TEMPLE DOT EDU] On Behalf Of STANLEY R. HORWITZ
>Sent: Friday, September 09, 2011 9:29 AM
>To: NETWORKER AT LISTSERV.TEMPLE DOT EDU
>Subject: [Networker] Basic questions about backing up Oracle databases
>
>Greetings everyone,
>
>We currently take a very basic approach to backing up our ERP 
>Oracle databases (all on Red Hat Linux). Each service (e.g., 
>Kronos, Cognos, payroll, student grades, etc.) runs its own 
>Oracle database on it's own server. Our Oracle DB 
>administrator has a cron job on each server that runs once 
>daily that does an RMAN dump to another Linux server who's 
>entire reason for being is to receive those RMAN backups. All 
>are Oracle 10.2g.
>
>The backup server that is a repository for these Oracle RMAN 
>backups is also a Linux box with it's storage on an EMC SAN 
>where each server's DB is backed up via RMAN to a different 
>directory on it's SAN mount point. I currently have a 
>regularly scheduled NetWorker group that backs up that one 
>(and only one) backup server once a day to a Data Domain 
>system. There is nothing fancy about this. In this group, I 
>use NetWorker 7.2.1's group cloning feature to clone the data 
>from our Data Domain system to encrypted LTO-5 tape over a 
>fiber channel link. The backup to Data Domain from the Oracle 
>RMAN backup server runs over a standard Gige network. Of 
>course, the cloning is a single stream of data by the time it 
>goes to tape. I also only have two tape drives to work with in 
>my tape library and about fifty application servers to back up 
>to Data Domain then group clone to tape in addition to this 
>Oracle data.
>
>This approach has been working out fairly well, but senior 
>management wants to reduce the backup window so we can get our 
>tapes off-site earlier in the day, which poses technical 
>challenges (no surprise there). Our tapes get sent off-site 
>late every afternoon and our DBA's RMAN backup scripts begin 
>at 5:00pm. If heh starts those scripts earlier in the day, we 
>will take a performance hit on our Oracle servers, which will 
>degrade performance for our ERP system's users.
>
>The solution I have in mind is to take the intermediary backup 
>server out of the equation and instead, let NetWorker send the 
>RMAN data directly from the Oracle databases to our Data 
>Domain system instead of the various cron scripts. This 
>approach is bound to be much faster than our current 
>simplistic approach, plus it will save several TB of capacity 
>on our SAN.
>
>For this approach, I am assuming, but I might be wrong, that 
>we would need to install the "NetWorker Module for Databases 
>and Applications." Unfortunately, the administrator guide for 
>that module is scant with information on how to set this up. 
>What I need to do, I think, is to install this module on each 
>of our ERP database servers. Right? Then the cron jobs we 
>currently have on each server would instead be scheduled via a 
>single group or would each one need to be scheduled as a 
>separate group? If separate groups are necessary, how do I 
>clone the sets of data to a tape drive in a single stream like 
>what I have now? I only have two tape drives. Unless 
>management wants to invest in more LTO-5 drives, I need the 
>other tape drive to clone data other than our ERP Oracle 
>databases. The NetWorker server is 64-bit Red Hat Linux with 
>the 64-bit NetWorker 7.2.1 Power Edition NetWorker server 
>software on it.
>
>Any comments or recommendations on which specific NetWorker 
>module to use and a high level approach to implement this are 
>more than welcome.
>
>To sign off this list, send email to 
>listserv AT listserv.temple DOT edu and type "signoff networker" in 
>the body of the email. Please write to 
>networker-request AT listserv.temple DOT edu if you have any problems 
>with this list. You can access the archives at 
>http://listserv.temple.edu/archives/networker.html or
>via RSS at http://listserv.temple.edu/cgi-bin/wa?RSS&L=NETWORKER
>

To sign off this list, send email to listserv AT listserv.temple DOT edu and 
type "signoff networker" in the body of the email. Please write to 
networker-request AT listserv.temple DOT edu if you have any problems with this 
list. You can access the archives at 
http://listserv.temple.edu/archives/networker.html or
via RSS at http://listserv.temple.edu/cgi-bin/wa?RSS&L=NETWORKER

<Prev in Thread] Current Thread [Next in Thread>