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: High availability of Oracle DB

Re: High availability of Oracle DB

From: Pete Sharman <peter.sharman_at_oracle.com>
Date: Thu, 17 Jan 2002 08:00:05 -0800
Message-ID: <owC18.12$nx3.78@inet-nntp1.oracle.com>


And that's exactly where OPS and RAC come in handy. But first let me clarify something that was stated earlier. OPS and RAC are not "needed only if you want several instances on each node access one database" but rather allow you to have multiple instances on DIFFERENT nodes accessing a shared database (I think this is what the original poster of that comment meant, but this wording just clarifies the point).

Coming back to the original point I wanted to make, OPS and RAC give you higher availability than a straight cluster failover can. If you're using cluster failover (whether that be using ServiceGuard, HACMP or whatever) you need to restart the database. With OPS and RAC, you don't need to do that. The instance is already running, and in most cases I've seen TAF used to fail over connections and even queries without re-entering the selects. For those sites that have really high availability needs, restarting the database just may not be an option. It's basically horses for courses, right? Pick the solution based on your needs. Either will work, you just need to determine the business need first.

--
HTH.  Additions and corrections welcome.

Pete
Author of "Oracle8i: Architecture and Administration Exam Cram"


"Controlling developers is like herding cats."
Kevin Loney, Oracle DBA Handbook
"Oh no, it's not. It's much harder than that!"
Bruce Pihlamae, long-term Oracle DBA
"Hieraklion" <hieraklion_at_spray.fr> wrote in message
news:3C46B58F.3000008_at_spray.fr...
"Bonjour" from Paris,
Stephan is right : all the sessions are lost (rollbacked) and the users or the applications need to be re-started or reconnected to the base. In fact the failover (i.e the launch of the instance on the second machine) is a shutdown immediate/startup standard procedure. Sometimes it's a shutdown abort/startup procedure, it depends of the nature of the problem requiring the failover. The longer is the switch of the disks/volume/and fs. I have no experience on HACMP but i use Veritas Cluster Server with Oracle on Sun Sparc E3500 cluster and this HA architecture is great ! No particular pbm during one year of production. And one failover due to a network problem - the first server had lost is IP due to a catalyst pbm).
"Au revoir" from Paris
Hieraklion Stefan wrote: Hi,Hieraklion is right you don't need ops. But - takeover your volumes takes its time- all the sessions are lost with TAF the can be reconnected or you connect to both instances- recovery time is shorterRegardsStefanHieraklion wrote:
"Bonjour" from Paris,In order to creatz a HA architecture, you don't need to
acquire OPS. OPSis needed only if you want several instances on each node access onedatabase.A HA architecture requires : - sharable disks (for the dbfiles) on the two nodes - a soft in order to manage the HA (descripbe the HA resources, the type of failer over ... I assume HACMdoes it ...) - virtual IP adresse for the listener ...."Au revoir" from ParisHieraklion James wrote: There are two Oracle databases reside on two RS6000 H80s. Each of themis running on a separeted machine. We want to build a HACMP for them,so in case of any server crash, two DBs are running on another machineat the same time.Is there any technical problems? Any things to be aware of (listner,port...)? We need Oracle Enterprise version or OPS.ThanksJames
Received on Thu Jan 17 2002 - 10:00:05 CST

Original text of this message

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