SQL2005 proactive approch monitoring alert

Question

SQL2005 proactive approch monitoring alert

SQL Galaxy

One Orange Chip

Points: 29472
More actions
September 20, 2010 at 4:39 am

#146970

Hi friends,
I want to be configure SQL agent for alert in SQL 2005, an alert there are three type available in SQLSERVER.
1.SQL Event alert – which is important event to be monitor and configure counter name & threshold values for Alert received
2.SQL performance Condition alert - which is important counter to be monitor and configure counter name & threshold values for Alert received
3. WMI event - which is important counter to be monitor and configure counter name & threshold values for Alert received
Please can you suggestion me, which are event should go for configuring and getting alert for Periodically monitoring?
Thanks
ananda

Viewing 4 posts - 1 through 3 (of 3 total)

You must be logged in to reply to this topic. Login to reply

Bhuvnesh SSC Guru Points: 59351 More actions · Answer 1

ananda.murugesan (9/20/2010)
2.SQL performance Condition alert - which is important counter to be monitor and configure counter name & threshold values for Alert received

Memory, CPU, IO usage.

-------Bhuvnesh----------
I work only to learn Sql Server...though my company pays me for getting their stuff done;-)

SQL Galaxy One Orange Chip Points: 29472 More actions · Answer 2

Hi Bhuvnesh thanks for reply, could you tell me how much Threshold precentage to be set in value column for those counter names.

regds

ananda

Keith Langmead Ten Centuries Points: 1048 More actions · Answer 3

Unless Bhuvnesh or anyone else has any specific guidance I would suggest this is a bit of a 'how long is a piece of string' question, in the that answer is that it depends. It depends very much on what is normal on your system. CPU is probably the easiest to judge, since obviously you want to be alerted if it is constantly maxed out at 100%, but beyond that, how far below that you would want to be alerted about depends on your system.

If the system is a busy one with regular high CPU load then you don't want to set the threshold too low otherwise you'll constantly be alerted, conversely if the system is normally very quiet, you might want to set it at 90%, or 80% or 70% or whatever else would indicate a potential issue.

I would suggest first monitoring the system at different times when there is likely to be different loads, and try to determine a general base line above which the server does go. Then configure the alerts in such a way that you can monitor them, without being bothered by them too much, and see which if any generate an alert. If you find you keep getting alerted then the threshold is probably too low, and you need to adjust it.

Getting too many alerts is almost as useless as getting none at all, since after a while you'll stop responding to them with the assumption that they're false positives, at which point a real issue will arise and be missed.