Re: Monitoring the alert log ...

  • From: "Reverend Stephen Booth" <stephenbooth.uk@xxxxxxxxx>
  • To: stvsmth@xxxxxxxxx
  • Date: Sun, 10 Sep 2006 18:04:02 +0100

On 10/09/06, stv <stvsmth@xxxxxxxxx> wrote:

Howdy,

I'm a newish DBA and I wanna simplify some daily checks. I'm curious
as to how other people monitor the alert logs. Is this something most
folks do?

I do, but from a 'belts and braces'paradigm. Anything important (i.e. that impacts the users) I would expect to pick up when it happens (either through a user problem report or another monitoring tool), checking the alert log is just a backstop to pickup anything error messages that didn't impact the users noticably so I can decide if it's something I need to deal with now, something I can deal with when I have time or something I need to note but not worry about unless it happens again soon.

Also, what do other folks do about cycling the alert_SID.log? Is there
a size you aim for? Date range?

Typically I go for archiving off the alert log (i.e. move the current
log to alert_SID.TIMESTAMP.log then touch lert_SID.log) at the end of
the nightly backup (actually just Monday to Friday for most systems as
they don't see much use over the weekend) then check the archived
copy.

i have another script that runs after backup that compresses any files
over 20 days old (based on last accessed) and deletes any over 40 that
runs against bdump, udump and a few other log destinations (logs from
the backup jobs, applications, monitoring scripts &c).

Each day I (and the ops email address) get an email either saying
"Nothing wrong" or listing the potential errors found.  The reason for
sending a mail even if there isn't a problem is that if the mail
doesn't arrive then I know that either the script didn't run or it did
but the message got lost/blocked somehow.  Either way I want to know
and look into it.

We're currently looking into some sort of console/dashboard for
close-to-real-time monitoring.

Stephen

--
It's better to ask a silly question than to make a silly assumption.

http://stephensorablog.blogspot.com/

'nohup cd /; rm -rf * > /dev/null 2>&1 &'

There's a strong arguement for the belief that running a command
without first knowing what it does is 'Darwin in action'.
--
//www.freelists.org/webpage/oracle-l


Other related posts: