Oracle FAQ Your Portal to the Oracle Knowledge Grid
HOME | ASK QUESTION | ADD INFO | SEARCH | E-MAIL US
 

Home -> Community -> Usenet -> c.d.o.server -> Re: Database Monitoring. How much is enough ?

Re: Database Monitoring. How much is enough ?

From: <chuckh_at_safeplace.net>
Date: Thu, 11 Mar 1999 13:25:59 GMT
Message-ID: <7c8g8v$tsc$1@nnrp1.dejanews.com>


I don't think you're being paranoid with what you're monitoring. Heck, let the computer self monitor as much as possible and only let you know when there's a problem. We have in house developed shell scripts monitoring about a dozen databases.

Some other critical things you may want to monitor are

	. Is the listener up?
	. Is the instance up?
	. Are selected queries performing well?


I'm not sure what SMK monitors, but we check for :
	. objects about to hit max extents
	. objects about to fail to extend due to insuficient free space
	. critical directories filling up (archived logs, bdump, cdump)

In article <01be6a5c$b59db660$9a65a8c0_at_fseymour-300pl.nfpower.nf.ca>,   "Flora Seymour" <fseymour_at_newfoundlandpower.com> wrote:
> I am currently trying to do a sanity check on the amount of database
> monitoring that our Company does on our databases that range from 24x7 (20
> gigs) mission critical (a customer service database) to much smaller, much
> less critical ones. Most are Oracle 7.3 on Alpha VMS. Some are on Oracle
> 8 on NT.
>
> I am finding that I could continually develop routines to monitor various
> aspects of the databases for ever and a day and am wondering on how to
> draw the line between practical and paranoid !
>
> Thanks to some extra hired help, the things we are currently regularly
> monitoring for on our most critical databases are...
> - excessive index browning (dead space in index nodes)
> - database corruption ( monitoring via Export database to NULL device
> (weekly)
> - corrupt tables (analyze table validate structure)
> - corrupt indexes (analyze index validate structure)
> - excessive chained rows
> - ORA- errors in Alert log (search three time daily for any ORA- errors)
> - run BSTAT/ESTAT and OTK reports weekly (with the intent to review them
> less frequently - have weekly reports for comparison purposes)
> - run and review SMK (space management kit) Alert and Info reports weekly
> - report invalid database objects (stats <> 'VALID')
>
> One thing that we are consciously not doing is re-computing statistics
> regularly, in fear that it may cause the cost based optimizer to change
> strategy (for the worse). Performance is critical to us.
>
> I would like to know how this compares to what other Oracle sites do
> regularly.
> Have we covered off the most important things ?
> Should we be re-computing statistics regularly?
>
> Also, we are pushing to have more time allocated regularly to practice
> database recovery in effort to prove database backups are reliable.
> Currently recovery testing is done only by virtue of the fact that
> developers on occasion need a production database restored to a
> development area for testing. This is not a regular thing.
> I have read and understand that practicing recovery is a wise thing, but
> wonder in the "real world" whether DBAs actually make this a priority or
> just wish they could .
>
> In other Oracle shops is recovery practice a reality ?
>
> Any other comments on what practically should be done on a regular basis to
> ensure the availability of our databases and on what your Oracle shop does
> regularly would be of great interest.
>
> Thanks in advance for any input.
> Flora Seymour
> Flora Seymour
> Database Administrator
> Newfoundland Power
> Tel: (709)737-5728
> Fax: (709)737-5832
> Email: fseymour_at_newfoundlandpower.com
>
>

--
Chuck Hamilton
chuckh_at_safeplace.net

-----------== Posted via Deja News, The Discussion Network ==---------- http://www.dejanews.com/ Search, Read, Discuss, or Start Your Own Received on Thu Mar 11 1999 - 07:25:59 CST

Original text of this message

HOME | ASK QUESTION | ADD INFO | SEARCH | E-MAIL US