About 10 months ago the same topic came along at another list. This was my answer (slightly edited for this list and this year): > I had an interview with Ken Jacobs, some years ago, and another one > September 2004, and both times asked him when Oracle will support > Rolling Upgrades. The first time he told me they were working on it, > and in September 2004 he elaborated on the subject and told me there > are rolling upgrades now. > > 'Almost', I answered. The fact that SQL-Apply means there are two > independent databases, in stead of 'clones' as in Physical Standby, > means that there is no Transparent Application Failover, or even > Translucent Application Failover (thank you Pete Sharman). It means > that the application has to be restarted/coalesced/interrupted > whatsoever (or whatever nice marketing speak you can invent to hide > the true meaning). So, there is NO rolling upgrade. Period. > Ken agreed. > > They get close, but there are so many pitfalls and tiny small > problems. What if the application relies on rowid's? They're different > in two independent databases. All there is the whole complexity of > maintainting a Logical Standby, with unsupported datatypes and > unsupported DML commands. There is the scaleability problem of > SQL-Apply. There is the need for supplemental logging on the primary. > If the whole RAC-setup is created with scaleability in mind, how to > cope with the single instance SQL-Apply part? It probably cannot keep > up with the rest of the system. Apart from that, it supports one-off > patches only, not patch sets (or has that changed already?). > > I do not follow the Meta Link approach of upgrading Data Guard > environments in their connected state. I use to defer the redo > forwarding while upgrading the primary. After the primary is patched > succesfully, and tested, and restarted, I reinstatiate the standby. > > Keeping the standby connected jeopardizes the whole conecpt of HA, in > my opinion. If the patch process fails for some reasons, you end up > with two unusable databases. My approach gives the possibility of > activating the standby immediately after the patch fails, while > postponing the patch process to a later moment, after more testing. > > Just my $0.02 > Best regards, Carel-Jan Engel === If you think education is expensive, try ignorance. (Derek Bok) === On Wed, 2006-03-29 at 11:21 -0800, oracle-l-bounce@xxxxxxxxxxxxx wrote: > For a while they were saying that a Shared Oracle Home > for RAC was a bad idea because "you can't do a rolling upgrade" > with a single shared Oracle Home. Then people started calling them > on the fact that you can't do a rolling upgrade AT ALL without > Data Guard involvement... > > > marketing > > > > >>>-----Original Message----- > >>>From: oracle-l-bounce@xxxxxxxxxxxxx > >>>[mailto:oracle-l-bounce@xxxxxxxxxxxxx] On Behalf Of Michael Fontana > >>>Sent: Wednesday, March 29, 2006 10:35 AM > >>>To: Jared Still; davidsharples@xxxxxxxxx > >>>Cc: Thomas.Mercadante@xxxxxxxxxxxxxxxxx; oracle-l@xxxxxxxxxxxxx > >>>Subject: RE: ZERO Database Downtime??? > >>> > >>>Excuse me, but "little or no" is not the same as "Zero Downtime". > >>> > >>>A logical SQL apply is also not feasible in a large > >>>installation, or if there is complexity, such as RAC. > -- > //www.freelists.org/webpage/oracle-l > >