Re: Questions about ORA-600
Date: Wed, 28 Jan 2015 12:57:37 -0500
Message-ID: <CAAaXtLDyj_R1ENOttpTx=ucJWRB_3P06bCzb8jx6FtYBAm+M4w_at_mail.gmail.com>
Users can ask some strange questions sometimes. This sounds almost like a case where a DBA was accused of incompetence during a staff meeting for having allowed the database to corrupt blocks on disk.
In this case, there's not a lot more you can say.
If a problem is reproducible, then you apply the patch, perform the steps to reproduce the error, and demonstrate that the problem is fixed.
If a problem is *not* reproducible, there is no way you can prove that it is fixed. Neither can you "prove" that it (still) exists. You can only "hope", and the "confidence" of that "hope" increases over time as the problem fails to re-appear.
As for the frequency of the ORA-00600 errors, a review of your alert-logs will (should) provide a history of exactly when the errors have occurred in recent history. If the user is keenly interested, you can answer the question by supplying a list of dates and times. Maybe even chart it in an Excel graph, or something. And then, if your organization supports it, charge back to the user's group/department for every minute of time spent. ;-)
Of course, you can only answer the question of how often a *particular* ORA-00600 error has occurred in your particular databases. How often they occur in general is impossible to answer, as most go unreported even to Oracle, and 99.999999% are certainly unreported to you. (If anybody out there is receiving reports of every ORA-00600 that happens worldwide, they have my deepest sympathy!)
On Wed, Jan 28, 2015 at 3:37 AM, Prem J <jprem_at_outlook.com> wrote:
> Friends ,
>
> Got below ORA-600 on one of our customer's 10.2.0.3 (sparc) box .
>
>
> *ORA-600 [KCBGET_37] / ORA-600 [1100] AND INSTANCE CRASH(not reproducible
> at will . and this bug is related to some buffer cache corruption)*
>
> So we suggested to apply patch 6110331 - and to monitor the instance for a
> week .
>
> Questions from user :
>
> 1) User : what's the frequency at which ora-600 error occurs ?
> Me : ora-600 is mostly due to bugs and can't really say when it will pop
> up . there is so clear cut way to say when it will happen again.
>
> 2) user : when you ora-600 are rare and happens at random - on what basis
> do you 1 week of monitoring the instance (after patch) will suffice ?
> me : usually - 1 week is the minimum time you monitor the instance -
> ensure that you test all your important business processes and if all goes
> thro' fine and if ora-600 does not happen again . then you should be safe .
>
> What would have been your reply for the above 2 questions ? I am looking
> for stronger/deeper points about ORA-600 occurrence and how it should be
> tested (after patch) . Would like to rephrase my next (to user) reply more
> technically & politically correct.
>
> Regards,
> JP
>
-- http://www.freelists.org/webpage/oracle-lReceived on Wed Jan 28 2015 - 18:57:37 CET