Oracle FAQ Your Portal to the Oracle Knowledge Grid
HOME | ASK QUESTION | ADD INFO | SEARCH | E-MAIL US
 

Home -> Community -> Usenet -> c.d.o.server -> Re: single node RAC system?

Re: single node RAC system?

From: DA Morgan <damorgan_at_psoug.org>
Date: Thu, 31 May 2007 18:00:11 -0700
Message-ID: <1180659609.791839@bubbleator.drizzle.com>


herta wrote:

> On Jun 1, 12:52 am, DA Morgan <damor..._at_psoug.org> wrote:
>> herta wrote:
>>> On May 29, 5:39 pm, DA Morgan <damor..._at_psoug.org> wrote:
>>>> herta wrote:
>>>>> On May 26, 2:25 am, DA Morgan <damor..._at_psoug.org> wrote:
>>>>>> herta wrote:

>>>>>>> We are currently setting up our disaster recovery site (DRS).
>>>>>>> For Oracle, we want to use Dataguard to copy our databases to physical
>>>>>>> standby databases on a single system at the DRS.
>>>>>>> The single system is quite capable of running the Dataguard load, but
>>>>>>> should we need to run production at the DSR, it will not suffice. We
>>>>>>> would therefore like to set up this system as single node cluster, so
>>>>>>> that we can easily add instances if necessary.
>>>>>>> I tried installing clusterware on a single node, but the start of ONS
>>>>>>> fails with:
>>>>>>> Start of `ora.l27.ons` on member `l27` failed.
>>>>>>> CRS-1006: No more members to consider
>>>>>>> CRS-0215: Could not start resource 'ora.l27.ons'.
>>>>>>> Metalink article 2411114.1 suggests that the set up is possible,
>>>>>>> though the environment described there differs from ours (we use
>>>>>>> 64-bit SLES 10, Oracle RAC 10.2.0.3, ASM).
>>>>>>> It also warns that "this configuration is not certified or supported
>>>>>>> by Oracle Support Services".
>>>>>>> Has anyone tried setting up a single node cluster? If so, did you get
>>>>>>> the ONS error? How did you work around it?
>>>>>>> As to the support status, in theory, we could set up a two node
>>>>>>> cluster, and then take one node down. How would that differ from a
>>>>>>> single node cluster configuration?
>>>>>>> Alternative suggestions to achieve our goals are also welcome, of
>>>>>>> course.
>>>>>>> FWIIW, I opened an SR with Oracle four days ago, but they have yet to
>>>>>>> respond with something other than "we're looking into it'. A previous
>>>>>>> posting in ORACLE_DBA_EXPERTS also remained unanswered.
>>>>>>> I'm sort of pressed for time as the goal is to ship this to the DRS in
>>>>>>> three
>>>>>>> weeks time, and just getting a first copy of the physical databases on
>>>>>>> the storage is going to take a while.
>>>>>>> Kind regards,
>>>>>>> Herta
>>>>>> Build a two-node cluster and then shut down one machine.
>>>>>> Suggestions that you use VMWare will result in an unsupported environment.
>>>>>> --
>>>>>> Daniel A. Morgan
>>>>>> University of Washington
>>>>>> damor..._at_x.washington.edu
>>>>>> (replace x with u to respond)
>>>>>> Puget Sound Oracle Users Groupwww.psoug.org
>>>>> Thanks, Daniel.    We're considering this, using a test system as our
>>>>> second member, but we're wondering what impact this would have when
>>>>> we'd need to upgrade the software.  (We cannot ship the test system to
>>>>> the DR site.)  Has anyone actually done this?
>>>>> Kind regards,
>>>>> Herta
>>>> We just got done doing this for a public company where they are
>>>> doing Data Guard from RAC cluster to single instance but want the
>>>> ability to quickly turn that single instance into a cluster by
>>>> adding nodes.
>>>> --
>>>> Daniel A. Morgan
>>>> University of Washington
>>>> damor..._at_x.washington.edu
>>>> (replace x with u to respond)
>>>> Puget Sound Oracle Users Groupwww.psoug.org
>>> Yes, I know I can set up a 2-node cluster, and then remove one node.
>>> I took the Oracle RAC course, and removing a system from the cluster
>>> and adding another one  was the last exercise of the training.
>>> I'm just worried about how we can upgrade that one node.
>>> As explained, we want to run Dataguard standby databases on this
>>> system, and as you may know, Dataguard requires the primary and
>>> standby databases to run the same software versions.  I would hate not
>>> being able to install a critical patch in production just because it
>>> cannot be installed on our DR system.
>>> Oracle has documented setting up a single node cluster in Oracle 9i
>>> without having to start from a two node cluster.
>>> My colleague read some Oracle 10g beta documentation with a similar
>>> setup, but the official documentation does not mention it anymore.
>>> I tried in vain to install a single node cluster, and the 'CRS-1006:
>>> No more members to consider' error has me worried that even when I'd
>>> start with a two node cluster and remove one node, the one node won't
>>> be upgradeable because it lacks other cluster members.
>>> Given the changes between 9i and 10g, I am reluctant to follow that
>>> path without confirmation from Oracle that they will support this
>>> configuration.
>>> (Still no word from them after 10 days, though. :-(  )
>>> Kind regards,
>>> Herta
>> I'm a bit confused by what is confusing you.
>>
>> A two-node RAC cluster is a special case. A cluster, to allow for
>> rolling upgrades, should be three or more nodes.
>> --
>> Daniel A. Morgan
>> University of Washington
>> damor..._at_x.washington.edu
>> (replace x with u to respond)
>> Puget Sound Oracle Users Groupwww.psoug.org
> 
> Hi Daniel,
> 
> I'm not even worried about rolling upgrades at this stage.  Though I
> didn't know that you couldn't do a rolling upgrade on a two-node
> cluster.  We just moved our production and non-production environments
> to two-node clusters, so this definitely is relevant info for us.  Do
> you remember where you read about this?  (If not, I'll hunt around
> myself, but I'm rather pressed for time right now.)

It isn't that you can't do a rolling upgrade. Rather that when you offline one node ... you lose TAF, lose FCF, lose load balancing, lose that which makes RAC RAC. You no longer have a cluster.

> About the initial issue: when I try to install a single node cluster,
> the error I get suggests that the installation script is looking for a
> second node,

Which is why I suggested creating a two node cluster and burning the second node to the ground as soon as you were finished or doing a shutdown immediate so it isn't live.

-- 
Daniel A. Morgan
University of Washington
damorgan_at_x.washington.edu
(replace x with u to respond)
Puget Sound Oracle Users Group
www.psoug.org
Received on Thu May 31 2007 - 20:00:11 CDT

Original text of this message

HOME | ASK QUESTION | ADD INFO | SEARCH | E-MAIL US