StefanKapitza <skapitza_at_volcanomail.com> wrote:
> On Jul 13, 8:48 am, Alexander Skwar <alexan..._at_skwar.name> wrote:
>> StefanKapitza <skapi..._at_volcanomail.com> wrote:
>> > On Jul 13, 8:21 am, Alexander Skwar <alexan..._at_skwar.name> wrote:
>> >> sybra..._at_hccnet.nl <sybra..._at_hccnet.nl> wrote:
>> >> > On Thu, 12 Jul 2007 23:35:14 +0200, Alexander Skwar
>> >> > <use..._at_alexander.skwar.name> wrote:
>> >> >>>>From where does RMAN/Oracle
>> >> >>>>pull the data about what has been done between 22:46 and 22:59,
>> >> >>>>if archive logs aren't available?
>>
>> >> > The answer to this question is of course: from nowhere.
>>
>> >> Thanks. So how's that an improvement over doing an EXP?
>> [...]
>> > The Improvment was (check your Topic) no Downtime.
>>
>> Well. But please check my OP,
>> <news:2877861.rzr6KgmIHE_at_kn.gn.rtr.message-center.info>, where I
>> explained that an RMAN backup would mean either having a larger downtime
>> or the odd chance of a user being able to modify something "in between".
>> Let me quote the relevant part:
>>
>> | I'm now thinking about how to fit RMAN into this picture. I think it
>> | might look something like this:
>> |
>> | 1) Shut down application, which uses Oracle as a backend
>> | 2) Have RMAN create backup of database
>> | 3) Create filesystem snapshots with ZFS on Solaris 10
>> | 4) Start backup to tape of filesystem snapshots. When done, remove
>> | snapshots 5) Startup application
>>
>> If I were just to do an EXP, I also would have no downtime, wouldn't I?
>> I mean, after all a EXP dump can be done while the database is still in
>> production.
>>
>> The way the application is made, I need to have a downtime, if I want
>> to make absolutely sure that the application is in a consistent state.
>> I don't need a downtime of Oracle, that's true. But I care much more
>> about the downtime of the application then about the downtime of Oracle -
>> especially if the Oracle downtime occurs at the same time, the
>> application is already down.
>>
>> But currently I'm not doing just an export. I'm doing cold backups and
>> this means a very /short/ downtime. At least in the way it's implemented
>> right now.
>>
>> | 1) Shut down application, which uses Oracle as a backend
>> | 2) Shut down Oracle
>> | 3) Create filesystem snapshots with ZFS on Solaris 10
>> | 4) Start backup to tape of filesystem snapshots. When done, remove
>> | snapshots 5) Startup Oracle
>> | 6) Startup application
>>
>> The advantage here is, that step 3 is very fast and that step 4 is done
>> in the background, while steps 5 and 6 are already started or completed.
>> As it is right now, I'm doing the exp right before I do step 1. I don't
>> care that much about the EXP being in sync with the application. My real
>> backup is the one I take from the snapshots. Those snapshots contain the
>> database in its shut down state - so it's a cold backup, as far as Oracle
>> is concerned. BTW: As the database is running in archivlog mode, I also
>> archived redo logs which I store as well on tape.
>>
>> Alexander Skwar
>
> Backup :
> - you take a snapshot a 23:00
=> Somebody modifies/deletes/adds a file
> - you start an Online Backup with Rman after that
>
> Restore :
> - restore your Filesystem Snapshot from 23:00
> - Recover your Database until time 23:00
>
> What is your Problem ?
Problem: Somebody could, at least in theory, come in between 23:00
and the time I start the RMAN backup and change something. That's
what I meant with "odd chance". Yes, that chance is very small, but
it's not 0. To get it to 0, I'd need to shutdown the application. Or
I need to make sure, that the two actions "create snapshot" and "start
RMAN backup" occur at the very same time.
Example: At 23:00, there are files "a", "b" & "c" in the vault directory.
At 23:30 (or 23:00:01, which doesn't make a difference, as we're talking
theory here), I start RMAN backup. Problem: At 23:15, somebody added
a file. This file is NOT part of the snapshot. The database, however,
says that this file should be present. But according to the filesystem,
it's not present.
To solve that problem, I need to make sure that no user can fiddle with
the files. The only way to do that, is to shut down the application.
Alexander Skwar
Received on Fri Jul 13 2007 - 02:34:41 CDT
