Thanks Hans. I completely agree and hence I do always follow it the way Oracle says. My point was, I thought (and have read) that Oracle says that GI at a higher version and DB at a lower version is supported but not the other way round (pls correct me if I am wrong here). Going by that, if opatch auto applies the PSU to DB home first, then it takes the DB to a higher version with the underlying GI at a lower version and that contradicted with Oracle's guidance, and hence my question. btw, as usual, the GI PSU in my context, does contain the DB PSU too. And I have observed the mentioned behaviour 100% of the times. Thanks On Thu, Feb 14, 2013 at 12:02 PM, Hans Forbrich <fuzzy.graybeard@xxxxxxxxx>wrote: > On 13/02/2013 8:42 AM, Purav Chovatia wrote: > > Hi, > > Anybody observed, that "opatch auto" first patches the DB home and then > the > > GI Home. Shouldn't it be the other way round? > > > > Thanks > As usual with Oracle: "It Depends". > > Patches MAY > - impact current software; > - impact current data dictionary; > - impact dependents > > Just because a GI patch is released, does not mean that it directly > impacts DBs that are involved with that GI. > > > If a GI patch does impact DB, you probably want to patch the DB first, > because you can add code to understand the new functionality as well as > the code to keep compatibility. If you upgraded the GI first, it might > introduce new functionality that the old DB code does not understand. > Which would mean the DB must be down from before the GI is patched to > after the DB is patched. > > Whereas patching DB first can (in theory, depending on the extent of the > GI patch impact, and whether we use RAC or only protected single node) > could mean the DB is down only for the DB patch, and perhaps a quick > reboot after the GI patch. > > > In fact, (bold statement here:) unless it impacts the interconnect > (specifically the way the DB rides over the interconnect) or the disk > (specifically the way the DB reads/writes ASM disks), I see limited > concern about the order in which things are patched. Does that mean we > can pick and choose the order when we do things manually? No - because > if we happen to choose wrong in production, it is on our neck, not > Oracle's. > > However, if Oracle does the patch testing correctly, they can decide > which combination of impacts are affected. > > Also, since Oracle has KSplice technology, and since that technology can > do in-place updates, I see no reason why Oracle could not do the equiv > to KSplice to GI and DB, at least on certain platforms, thereby patching > without ANY downtime at all. Which introduces a whole new set of > possibilities (and a whole new set of problems). > > > In conclusion: Do it the way Oracle says. If it doesn't work, we can > complain to Support. (LOL) > > /Hans > -- > //www.freelists.org/webpage/oracle-l > > > -- //www.freelists.org/webpage/oracle-l