Re: Cron management...

From: Hans Forbrich <fuzzy.graybeard_at_gmail.com>
Date: Wed, 22 Apr 2015 17:13:20 -0600
Message-ID: <55382B10.4000502_at_gmail.com>



On 22/04/2015 4:38 PM, Mladen Gogala (Redacted sender mgogala_at_yahoo.com for DMARC) wrote:
> On 04/22/2015 11:06 AM, Hans Forbrich wrote:
>>
>> Issue here is the definition of BackUp. Unfortunately, I have run
>> across too many people who believe the RMAN executable actually does
>> the work of backing up the database. My post was to allow access
>> official references on the process.
>>
>> As for central 'back up', the referenced doc mentions Oracle Secure
>> Backup, which is also a centralized mechanism similar in nature to
>> other commercial products. From that perspective, it is useful to
>> look at Oracle's implementation of what you describe, and that is
>> available at
>> http://docs.oracle.com/cd/E16926_01/doc.121/e16564/osb_rman_backup.htm#OBADM187
>>
> Hans, how does rman work? What happens when you do "allocate channel
> c0 device type SBT"? Several things happen:
>
> * oracle process is allocated.
> * MML library is mapped into the rman address space
> * sbtinit routine is executed to initialize the MML side of the channel.
>
> After that, when a "backup" command is issued, Oracle process reads
> the database and delivers the data to the channel, possibly using
> Oracle*Net. The channel then pumps the data toward the backup
> software. That's the nature of the beast. The "rman" executable is a
> client, similar to sqlplus, which executes the internal functions,
> built into the Oracle executable itself and delivers the data to the
> destination using the channels. If the channels are defined using MML
> library ("device type SBT"), then the sbtwrite and sbtwrite2 routines
> are used to send the data to the backup software, be it NB, Avamar,
> Simpana or something else.
> In other words, if there is a centralized node used to backup all the
> database by establishing remote connection to the database and opening
> channel toward backup software, that node is a critical point of failure.
>

Glad we agree.
/Hans
--
http://www.freelists.org/webpage/oracle-l
Received on Thu Apr 23 2015 - 01:13:20 CEST

Original text of this message