Re: The penny hasn't dropped yet...
Date: Wed, 3 Mar 2010 09:58:02 -0800 (PST)
Message-ID: <893f32d5-6cfa-46ca-934b-8842ab41cdce_at_a10g2000pri.googlegroups.com>
On Mar 3, 4:01 am, Noons <wizofo..._at_yahoo.com.au> wrote:
> vsevolod afanassiev wrote,on my timestamp of 3/03/2010 8:15 PM:
>
> > I think it depends on the specifics of the case. If to investigate
> > this case Oracle Support needs init.ora parametes
> > or server patch level or similar then this information should be
> > provided in one form or another.
>
> Good. All they have to do is ask for it, if/when needed. Period.
I strongly agree with the sentiments in this thread. But, here I have to point out how common it is, to the point of being the mode, that people won't supply the necessary information up front (right Sybrand? :-) . From that viewpoint, it merely becomes an issue of degree as to how much they have to supply to initiate an SR. I'm sure we've all seen order entry systems that create a header as a separate transaction from the order lines - that's wrong from a design standpoint, but right from the viewpoint of having a resumable entry process (or points up the deficiency of not having multiple levels of transactions). So in an SR, they have to ask for certain information, it's needed at the start. It's not a big jump from there to analyzing the patterns of previous SR's to see what in most cases was needed. Where they cross the line is grabbing all information ever possibly needed. Think of this: if you have everyone upload all their trace files, you can mine that to create a decision support tool that can automate much trace file analysis. I'm sure you can name popular and not-so-popular people who have done that. So why not generalize that? Well, obviously that upsets everyone who wouldn't willingly upload all their data. Through the magic of Computer Aided Mass Hysteria, it becomes a potential publicity nightmare. So maybe we should convert potential to actual and tell Warticki about what we think on the MOS community :-D
>
> > For example if I ask Oracle Support to investigate ORA-04031 errors
> > then I would provide them with
> > - init.ora parameters
> > - contents of V$SGASTAT, STATS$SGASTAT
> > - if this is 10g and ASMM is enabled then information from the views
> > that show how memory was resized
> > - ORA-04031 trace files
> > - alert.log
>
> Exactly.
>
> > So may be they are asking for information that should have been
> > provided from the beginning
>
> Actually, they asked for nothing: they just sent that email as an auto follow-up
> to opening the SR, long before anyone had a look at it.
>
> > However if they are using lack of OCM to justify a delay in the
> > investigation when all necessary information has been provided then it
> > is bad.
>
> ORA-600 on a select from subpartitioned table, trace dump of simplified
> statement provided. What possible need would there be to know the MAC address
> of the network card in that system?
Well, what if that ORA-600 is hidden under several levels of technology stack? Way before you get there, you may indeed need to follow the problem including over the network. I would think most support calls are of the form "my program isn't working," I'm sure ora-600's are pretty scarce in the overall scheme of Oracle support. We would all like a direct line into tertiary level support, and personally I've noticed if the magic words are there, I often wind up there quick. But if I'm getting a java virtual error, I'm just as newbie as "my program isn't working" and I think an RDA is probably appropriate. OCM is just a proactive version of that, isn't it? The Network Is The Computer. And OCM puts all your Oracle usernames into a world readable file. Sigh.
jg
-- _at_home.com is bogus. http://twitter.com/safetyReceived on Wed Mar 03 2010 - 11:58:02 CST