System Health Check metrix

[YLTan]

Does anyone have a good sets of system performance metrix use in Perf.Manager for doing routine System Health Check? Basically, what I need is to know if my servers are not overloaded or running in critical resources shortage or some processes had gone wild, etc..

I am using HP SuperDome and on HPUX11i

[Michael Steele_2]

In my daily crons I'll have one graph for global CPU bottlenecks and one for global disk bottlenecks but there are also global process, global application, global memory, etc., etc. Since every sever is different you'll have to factor in your needs for yourself. These two however I have on every server.

Refer to the logglob measureware file to extract global heuristical data. Review to the measureware dictionary for bottleneck definitions. For example, a CPU bottleneck in measureware is the sum of three metrics: CPU utilization, higher than usual number of processes, and one or more jobs waiting in the run queue. If you export this collected data out to a graph, you can include the high water marks and make a nice bit of mgt. eye candy.

There are maybe 20 variables in logglob to choose from. Choose no more than six or you'll run out of column space on your graph.

I do have secondary extractions from logdev, logtran, etc., and occasionally I'll use them to investigate an outage that wasn't captured in real time. That's one of the nice things about these log files, the data is there for quite awhile and doesn't have to be captured in real time.

Management has always appreciated this.