On Fri, 16 Jun 2000, Michal Kara wrote:
> ...
> When I add it, you will be able to run:
> pcpmon -a day1=... -a day2=... -s -1d -a day3=... -s -2d
>
> Then, if you create yourself config which will contain expression:
> (day1:kernel.all.cpu.load+day2:kernel.all.cpu.load+day3:kernel.all.cpu.load)/3
>
> This would print average CPU load from the three specified days.
You may wish to review pmlogsummary. For all sorts of statistical
summaries from archives, we've found this to be a generally useful
tool.
I think one needs to be careful about overloading PCPMON ... our
experience has been that a number of smaller, focussed tools are more
effective than the swiss army knife style of all singing all dancing
tool. This is especially true when one moves from interactive
monitoring to the broader field of performance management.
> What I am doing right now is to add possibility of entering the expression
> and highlight specified parts of the graph (i.e., when CPU load was >10 etc.).
Another alternative may be to use pmie to filter the archive (pmie
already has very powerful rule evaluation features), and customize the
"notification" of the rules to produce timestamped events when the
rules are true, then have your tool read these notifications and
translate this into a visual alarms on the graph window.
|