(and I’m writing this from a positive perspective)
Now that the community is over the initial ‘Look over, there! It’s 11g!!!!’, there are some genuinely interesting blogs kicking around. (Maybe we need to give these things a little time?) I’d better only talk about a couple, though, or Dave Edwards will start telling me off for pre-empting this week’s Log Buffer!
For my part, my feelings about 11g haven’t changed from those I expressed to a fairly well known Oracle employee at last year’s UKOUG conference. I couldn’t see why people were so negative about the announcements in 2006. What were people expecting? Don’t they care about minor incremental improvements, features to improve test environment management and the like? (I’m just waiting for Noons to pop up and request some bug fixes here!)
I don’t expect to see any dramatic changes in such a mature product, but there are shed-loads of tiny little, ‘Oh, I could have done with that in the past’ improvements. I’m sad that way – I get excited about database-level default and temporary tablespaces and the like 😉 Getting the existing features to work the way I need them to is always a good starting point. In fact, Niall’s list on this forum posting rang true with the way I see the world. My list wouldn’t be the same, but there would be some UNION there. (Damn, I try to avoid database jokes like the plague normally!)
Anyway, others have taken more time to play with 11g and Howard’s started a series on new initialisation parameters which I’m looking forward to working my way through. Julian Dyke has also published some summary information on his website.
There’s quite a bit of debate about what precisely the CONTROL_MANAGEMENT_PACK_ACCESS parameter actually does. The MMON process (to use one example) is still running, but I suppose someone would need to trace it to establish whether there’s any resulting overhead. The default value of “DIAGNOSTIC+TUNING” seems strange to me, though. Even as someone who has argued that the minority who license these options is significant, I don’t doubt for a second that it’s still a minority and is likely to remain so. Based on that, it would have been far safer to have the default set to “NONE” to avoid anyone accidentally violating their license agreements and, on that subject, I see Mark Brinsmead’s had some response from Oracle.
My highlight of the first few days, though, is a discussion over on Oracle-L about the move back towards an OFA-compliant directory structure. I never truly got to the bottom of the muddled 10g approach and assumed I was missing something, but the fact that Oracle have reverted to the traditional approach might indicate that they were missing something! 😉 Yippee!
Either way, it’s all more material for my OFA-related presentation at UKOUG this year! (See, told you I had very boring tastes …)
Oh, and there are tons of excellent 11g technical blogs over at Amis and Pythian, but I’ll leave it there for now.
Nice roundup, Doug.
I’m still eagerly looking for bleeding edge candidates to enter my ‘Oracle 11g version banner’ (preferably on an esoteric flavour of Linux or a Commodore 64) competition.
Seriously though, most of my (Siebel) customers are still on Oracle 9i and are slowly and reluctantly pondering the implications, resources and risks of the migration to 10g.
I suspect most of the creases will have been ironed out (and/or I will be dead) by the time I encounter Oracle 11g in real-life production environments.
We’re finally going to patch our 9i databases to 9.2.0.8, but only because we were forced into it for support.
Still, at least all of the 8i are 8.1.7.4
Don’t ask …
8i, 9i….
bugger! I’m almost ashamed to admit we’re looking at 10.2.0.3.
Then again: after the last 10.2.0.2 inglorious crash and burn, I felt like reloading 8.1.7.4: at least it was stable.
But even with 10.2.0.3, we already have to patch it for Peoplesoft!
And you want me to look at 11g?
Yeah! Right…
On a good note: I’ve managed to convince management here that we need a “play” system. So there is a relatively old ex-web server Solaris box on its way back to the rack. It’s gonna have 11g in it, that’s for sure. And Apex.
If I can only convince them to get us another AIX box to dump all the release 7 Wintel databases floating around so I can slowly upgrade them to 10g…
OFA and layout for Oracle installs is certainly getting confusing.
I’ve always set ORACLE_BASE to be the same as the home directory of the oracle user. Now the installer recommends against it – so where should the oracle users home directory be ?
And now in 11g the installer put all the dumps etc under ORACLE_BASE/diag. What happened to $ORACLE_BASE/admin/$ORACLE_SID/… ?
I liked the old way 7/8i/9i I understood it. The 10g/11g changes just don’t seem logical to me. So looking forward to the you explaining in a presentation.
Andy
hey you didn’t used to work in a local authority did you? I have always gone with
$HOME = /home/oracle
$ORACLE_BASE = /x01/oracle (I never liked the app bit)
but here we have multiple Oracle accounts on a dev box all with $HOME==$ORACLE_BASE as you describe.
Getting EM to monitor all of these is – er challenging – unless you like changing perms on home directories.
Of course there’s always
c:\documents and settings\oracle
and
c:\program files\oracle
if you want to kill yourself 🙂
(I never liked the app bit)
Yeah, I know everyone hates that bit and thinks it’s a redundant annoyance, but OFA needs to be seen in the context of Oracle not being the only thing running on a system. Sometimes it has more flexibility than people might need, but that’s what I like about it.
Interestingly, I virtually forced a Unix Sys Admin to read OFA earlier this year and he loved it. He could understand the benefits perhaps more than most DBAs because he has to think of wider system management issues.
But I always think that anything as complex as a generic system configuration standard implicitly contains *lots* of room for debate!
I’ve always set ORACLE_BASE to be the same as the home directory of the oracle user. Now the installer recommends against it – so where should the oracle users home directory be ?
Well, I’ve seen ORACLE_BASE set to a different directory to the oracle user’s home directory, I suspect mainly to keep sys admins happy with a stub login directory that matches the other user accounts. It’s workable once the profile’s configured, but if you look at the most recent edition of the OFA paper on the Hotsos website, here’s what it has to say.
All Oracle software and
administrative data reside in subtrees of the Oracle Server owner’s login home directory. The name of this directory is the value to which Oracle advises that you set the environment variable ORACLE_BASE.
So there are definitely variations of this floating around!
And now in 11g the installer put all the dumps etc under ORACLE_BASE/diag. What happened to $ORACLE_BASE/admin/$ORACLE_SID/… ?
Oh, yeah, I forgot to mention that one! I noticed it on Kevin Closson’s blog and immediately thought ‘I wonder what that’s going to break?’
Looking at the reasons, I suspect it’s to do with wanting everyone to use Grid Control, but I need to think about it more.
I liked the old way 7/8i/9i I understood it. The 10g/11g changes just don’t seem logical to me. So looking forward to the you explaining in a presentation.
All in all, you can see why I want to talk about it. Will you be coming then? 😉