Metrics and Monitoring round table - Oct 30

[Jennifer Davis]

For our discussion we will be focusing on a single OS (Unix).  We'll 
determine the specific OS by attendees expertise.  Next month's round 
table will be on performance tuning, so this will specifically be about 
monitoring/metrics.  I'm sending some of these questions to the group as a 
whole to start some discussions about monitoring/metrics.

Why do you monitor?  Do you have different consumers to your monitoring 
solution? (execs, customers, oncall staff)  Who do you monitor for?
How do you choose what to monitor?  How do you not affect what you are 
monitoring?

In your role, can you influence the developers of the application you are 
supporting (is it inhouse software or external)?

How do you prevent bad activity from bringing down the service through 
monitoring?

Are there standards? best practices? Any recommended books?

What tools are available?

disk, memory, cpu, network - Does virtualization change anything?

Is there a standard way of breaking down an application to monitor it? 
What methods are in place to automate the monitoring structure?  How do 
you monitor a service that is provided to an external customer, and 
provide the customer with a mechanism to view the monitoring?

What is page-worthy?  Critical vs simple error?  When should something 
wake you up, or disturb weekend activities?


Jennifer