Re: alter database suspend

From: Volker Hetzer <firstname.lastname_at_ieee.org>
Date: Wed, 18 Oct 2006 15:32:18 +0200
Message-ID: <eh5ad2$j9l$1@nntp.fujitsu-siemens.com>

hpuxrac schrieb:
> Volker Hetzer wrote:

>> Hi!
>> Can anyone tell me at which point the i/o activity gets suspended?
>> For instance if a suspend happens
>>   - when only one out of three or four controlfiles has been written
>>   - ditto for redo mirrored redo logs
>>   - ARCn is in the middle of archiving a nice big 512MB redo log,
>> will those writes be completed?
>>

>
> Maybe it would be more helpful if you give us some background.
>
> Are you using suspend in some of your current processing? Do you have
> concerns about why it was developed and how it has been placed?
I'm just curious.
We use 10g on Linux, one copy of the archivelogs on our netapp filer (nfs). Last month we had a controller crash on the database (we could recover complety) and since then we think of mirroring some stuff on a netapp filer. We started out with the archivelogs, no snapshots, to get a feel for it.
The question for me is, can we make use of the snapshotting capabilities in some way?
I guess, performance aside, it would make sense to have a controlfile on the filer, on a separate volume. In case of a crash we could then decide to use the current or a previous controlfile. The same reasoning applies to a lesser extent to archived logs, even if it is unlikely that arc messes up logs already archived.

Since oracle's backup mode is for tablespaces I guess that suspend is for the other files and I'm wondering about the granularity of it.
Will suspend guarantee a completely written controlfile? Ditto for archived logs?

We have no pressing need, since rman works nicely for us. But I'd still like to understand it because it always pays to know ones options.

Lots of Greetings!
Volker

-- 
For email replies, please substitute the obvious.

Received on Wed Oct 18 2006 - 08:32:18 CDT