2006/12/18

wishlists

Howard Rogers blogged recently about a simple feature that would greatly simplify his Oracle dba tasks. In his case if roles worked properly in Oracle and/or user authentication methods were not linked to the user name itself, life would have been a lot easier.

Yes, there are heaps of these small enhancements that were requested from Oracle over the years. Very few made it to the latest versions of Oracle. If ever. It's a pity, really.

Historically, Oracle has always relied more on new features than actually making life easy for their current users/administrators. I could go on with a long list of requested enhancements that was never acted upon.

But quite frankly: I stopped bothering a long while ago! More or less when they decided that the solution to a skills shortage was to eliminate the skill. The problem of course is they tried to eliminate the skill of using a brain...



While I'm in rant mode, I might as well mention the fixation that some folks have with automating Oracle startups and shutdowns with those of the operating system itself.


Folks, the VERY LAST thing you want done to a database service - and I don't care if this is SQL Server, db2 or Oracle - is to have it start up automatically with the OS. It's a sure fire recipe for a potentially terminally corrupt database: all you need is for the last shutdown to have been caused by a hardware crash involving the right kind of loss of disk. And no, RAC/SAN/Clustering won't help you here!

In the particular case of Oracle, you now will have ALL your multiplexed controlfile copies pointing to a failed, attempted recovery: the very last place you want to be at!




Of course: had Oracle paid attention to the many requests for enhancements, there were a few asking for controlfile mux copies to be offlined/onlined dynamically. That way folks could always turn off one of the muxed copies at system startup by default, then add it later manually if the db started up fine. But then again: why listen to folks who have been suffering - sorry: using - their product for decades? All "old Cobol dinosaurs", are they?...





Anyways....


Been busy last week with some minor personal and friend's problems: an old friend of mine lost his Mum to a combination of old age and cancer. That really touched me as I'm due for the same at some early future stage. I spent some time comforting him and thinking about the whole episode. Things appear to be better now so hopefully I'll have free time to continue my hobbies.


The next two weeks I'm on early-start shifts: that means I'll be at work at sparrow-fart and back home at around 4pm. I like that. Means around 4 hours of daylight in Sumer to catch up with all the little jobs that need doing and still have time to do some photos. Or go out for a quick 9-hole golf sortie. Or whatever. A much better way of running one's day. Wish I could make it permanent. But the other folks I share the work shifts with think the same way. (sniff...)


A word to the folks from the UK who might read this:
sorry about the Ashes.

But they were due back. There, I said it! (flak jacket on)


Catchyalata, folks.

5 Comments:

Anonymous Kevin Closson said...

"While I'm in rant mode, I might as well mention the fixation that some folks have with automating Oracle startups and shutdowns with those of the operating system itself."

...You aren't alone on that one, Noons, but I'm afraid this more than a fixation. It is nearly religion. I personally don't think the database should start up until "something" has validated underlying resources are all online. I have my opinion about how best to do that, but I'm not here to shill :-)

Sorry about your friend's loss too...

Thursday, December 21, 2006 3:53:00 am  
Blogger Noons said...

by all means, Kevin: blog away on that. and if some1 calls it a "shill", I'll send "the boys" around!

there are some truly bad "best practices" out there. this "startup automation" is one of the most blithering imbecile ones.

and very few people alerting folks to the real problems they may be facing if they blindly follow such "practices".

I'll blog here in future on the ones I really hate. but every bit of information helps.

contrary to most Oracle-sponsored "advisers", I tend to not blame the dbas for these things: people are only as good as the information they are given to work with. it's not their fault if they are fed a bumsteer to start with.

dbas out there need sound advice. instead of more "rules of thumb" and other such "best practices".

(reminds me I must blog on the "ebay uses j2ee, so it can't be shite" mantra)

Thursday, December 21, 2006 7:46:00 am  
Blogger Alex Gorbachev said...

Well, I don't agree that auto startup is a bad idea. In some environments this is a must as a delay due to human intervention means minutes (tens of minutes) of downtime.

Of course, it should be implemented properly and all required resources must be started/available prior to database startup. But hey, ain't those nice third party clusterware packages out there for that? I have excellent experience doing that with Veritas and I must say it works like a charm.

Sorry Kevin, I don't have any experience with your beloved one ;) but I'm sure it must have all required features to do the same.

Friday, December 22, 2006 2:57:00 pm  
Blogger Alex Gorbachev said...

Mate, do you have any way to enable comments feed on Blogger? For some blogs I like comments as much as posts and I subscribe to them. Like Kevin's blog. ;)

Friday, December 22, 2006 3:06:00 pm  
Blogger Noons said...

Sure, given a requirement for HA and the adequate hardware and software combo to make it happen, yes: it's a go.

However what I've found is in the vast majority of cases the database is simply re-started. Not a single confidence or hardware or file system access check. Not even a check if the controlfiles, redo logs and basic tablespaces are still there. Let's not even mention a simple "tail -50" of the alert log looking for any "ORA-" stuff.

Narh. Just a plain and simple entry to start the db. Bang. Recipe for potential disaster. Of course: 99 times out of a hundred, all will be well. Then the odd one out will cause an outage of an entire day. Or worse...


Dunno about the feeds: the new blogger is supposed to have them for comments. But I can't get the blog converted: their site is incredibly busy. Once it tappers down I'll give it another try.

Saturday, December 23, 2006 12:44:00 am  

Post a Comment

Links to this post:

Create a Link

<< Home