Re: More than one RAC instance/database on the same machine ?

"Niall Litchfield" <niall.litchfield_at_dial.pipex.com> wrote in message news:4184b263$0$1821$cc9e4d1f_at_news-text.dial.pipex.com...
> "Le JeanMimi" <scjm_at_noos.fr> wrote in message
> news:c7be5048.0410301212.172d66f8_at_posting.google.com...
>> First, thank you for your answers.
>>
>> For us, the point is availability (and not scalability).
>
>> If the machine 1 crashes, then all the work being done by A1 and B1
>> failover to machine 2 (TAF) and everything is very fast ... and
>> transparent.
>>
>> Does it make sense ?
>
> That is the marketing story yes. You should be aware though that in
> reality TAF requires your app to be written to be aware of it

Come on. TAF works fine so long as you are using OCI.

That does not require application awareness.

> and that not all statements (especially DDL) can failover.

Either I'm having a particularly grumpy day (always possible), or people are just posting silliness in abundance. TAF does not failover DML or DDL. It's purely select statements only. And no session variables or anything like that. This much has been known and written about since year dot, and the Oracle documentation is abundantly clear on the topic, so this should not be a surprise to anyone. I don't understand the "especially DDL" comment, since whatever applies to DDL in respect of TAF also applies in equal measure to DML. There's no "especially" about it.

Now. Were you trying to say that TAF is not all it's marketed up to be? That it only works for simple select statements? Good, because that's the undoubted truth.

Does TAF require the application to know about the failover of select statements? Of course not... otherwise it wouldn't be called "REAL APPLICATION clusters" would it?

I dunno. Maybe you're just trying to put some realism into the 'high availiability' claims. Fair enough... but don't denegrate TAF on unfair grounds. TAF has never claimed to be able to fail over DDL or DML. It is what it is, and what it is is readily understandable from a quick read of the doco.

> TAF emphatically does not mean that you can take an existing application
> and run it on a cluster and have it be magically node failure proof.

Yes it does. For select statements, that's precisely what it means.

Is there some perspective issue here? I don't expect TAF to do anything except deal with select statements. But within that particular goldfish bowl, it does what it does very well, and without the application needing to be aware of it. OCI is the only requirement, and no application re-writes necessary.

If your point is that the original poster seems to have a halo-induced view of TAF, and you are trying to bring him or her back down to Earth, so be it. And fair enough, I suppose. But I don't read that sort of naiive hopefulness into his comments.

In any case, let's try and just stick to the facts: TAF requires no application awareness. It works because OCI is in the picture somewhere. It fails over select statements only. No DML or DDL. And to resume a select statement on failover, there must be a load placed on the client for all select statements -which might be unacceptable. However, there are degress of TAFness, such that 'SELECT' is not the only method available.

Regards
HJR Received on Sun Oct 31 2004 - 05:10:54 CST