RE: Active/Passive "high availability"

  • From: "Michael Fontana" <MFontana@xxxxxxxxx>
  • To: <peter.schauss@xxxxxxx>, <oracle-l@xxxxxxxxxxxxx>
  • Date: Wed, 11 Apr 2007 18:11:06 -0400

If he is using a supported file system clustering solution from Oracle,
such as Sun or Veritas, this architecture COULD work (and these are not
cheap, and they are not simple to implement).  

We actually have built some similar systems for our hosting customers
running 8.1.7.  The only advantage you get is if the primary box is
damaged or corrupted in some way, you can "failover" to the alternate.
We have had minimal success with automating the failover process,
however.  A lot of times, whatever time savings you might have had gets
eaten up in analyzing the problem and deciding to failover or simply
restart after the root cause is resolved.  Because of the limitations
and ancient nature of this architecture, you'd be much better off trying
to convince him to use a standby solution from Oracle.  It is simple,
effective, supported by the vendor, and accomplishes nearly the same
thing, albeit with possible additional disk cost.

And oh yeah, if you have Oracle software on two boxes, whether you run
them or not, be prepared to negotiate payment for two licenses.  

And yes, you would need two separate password files, though I doubt you
could corrupt the data if the second instance started up, it would fail.


If your boss is like mine, when vague solutions are suggested, I just
pretend to be interested and wait to see if he formalizes the request.
Chances are, unless someone is driving him to this solution, you'll hear
nothing more about it.  But still, the standby solution is something you
ought to be recommending yourself, if you don't already have one.  It's
your job!



-----Original Message-----
From: oracle-l-bounce@xxxxxxxxxxxxx
[mailto:oracle-l-bounce@xxxxxxxxxxxxx] On Behalf Of Schauss, Peter
Sent: Wednesday, April 11, 2007 3:25 PM
To: oracle-l@xxxxxxxxxxxxx
Subject: Active/Passive "high availability"

My manager is proposing to cluster two Solaris 5.9 servers in order to
create a "high availability" solution.  At this stage he is a bit
unclear as to what software he would be using for the cluster, but he
does not envision purchasing any additional licenses from Oracle.  Our
Oracle version is 8.1.7.4.

His idea is to have an Oracle instance running on one box and a second
Oracle home on the second box, but not running.  In the event of a
failure on the active box, we would start the Oracle instance on the
other one to minimize down time.

Aside from my questions about what this configuration actually
accomplishes, my concerns would be:

-  Assuming that all of the data files are on the shared disk, would
this approach actually work?  

-  The password file is normally stored in $ORACLE_HOME/dbs.  If we have
two of them do they need to be synchronized?  Is there anything else in
the dbs directory which needs to be synchronized?

-  Would we risk corrupting our data if we accidentally started the
second instance?

-  Are there any Oracle licensing issues?

Thanks,
Peter Schauss
--
//www.freelists.org/webpage/oracle-l


--
//www.freelists.org/webpage/oracle-l


Other related posts: