Message-Id: <25968.338591@fatcity.com> From: "Tanel Poder" Date: Sun, 20 Jul 2003 16:52:56 +0300 Subject: Re: RAC or Oracle Fail Safe This is a multi-part message in MIME format. ------=_NextPart_000_00C8_01C34EDF.5EB24200 Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: quoted-printable RE: RAC or Oracle Fail Safe Hi! Of course you have to plan your servers' capacity accordingly, that in = the event of node failure, the other node will not get too loaded. When = one node crashes, the second one has to deal with queries and = transactions of both servers and we must not forget, that also rollback = of failed nodes uncommitted transactions has to be done! Of course when you got more nodes, then the impact of one node failure = would be smaller.. Tanel. Here I would beg to differ. RAC is more safe. _Almost_ half of your = online user (who are on surviving node) won't even notice the node = failure when one node in a RAC environment goes down. And those who were = unlucky users (connected to failed node), can connect immediately to the = surviving node, without any delay. Of course you need to configure these = parameters manually. While in OFS environment your users have to wait = till the time surviving node brings up the database and all related = services (listener etc) completely.=20 The price you may for these features is more money and more complex = database environment to manage. =20 ------=_NextPart_000_00C8_01C34EDF.5EB24200 Content-Type: text/html; charset="iso-8859-1" Content-Transfer-Encoding: quoted-printable RE: RAC or Oracle = Fail Safe

Hi!
Of=20 course you have to plan your servers' capacity accordingly, that in the = event of=20 node failure, the other node will not get too loaded. When one node = crashes, the=20 second one has to deal with queries and transactions of both servers and = we must=20 not forget, that also rollback of failed nodes uncommitted transactions = has to=20 be done!
 
Of=20 course when you got more nodes, then the impact of one node failure = would be=20 smaller..
 
Tanel.
 

Here I=20 would beg to differ. RAC is more safe. _Almost_ half of your online = user (who=20 are on surviving node) won=92t even notice the node failure when one = node in a=20 RAC environment goes down. And those who were unlucky users (connected = to=20 failed node), can connect immediately to the surviving node, without = any=20 delay. Of course you need to configure these parameters manually. = While in OFS=20 environment your users have to wait till the time surviving node = brings up the=20 database and all related services (listener etc) completely.=20

The=20 price you may for these features is more money and more complex = database=20 environment to manage.