guy keren wrote:
>
> On Fri, 1 Sep 2000, Aviram Jenik wrote:
>
> > Can anybody recommend a good way to really check my server load? I'm
> > currently checking the load using uptime (and check the average load
> > values), but this hardly seems a good way to really know whether the server
> > is 'sweating'.
> > Can anybody recommend any better way?
>
> you should be collecting info about CPU usage, disk IO, page in/out
> operations, network traffic, etc. then make graphs of all of them, and
> make this data visible some way. MRTG could be a good tool for visualizing
> this data, provided that you could write the extensins to track resources
> that it cannot track out-of-the-box.
>
> no single parameter would tell you the whole truth.
>
> to see disk ip and paging, use 'vmstat' (other unices also have 'iostat'
> and 'sar'). note about vmstat - the first line of output is usually
> meaningless. running vmstat in a mannaer such as 'vmstat 60' will produce
> an output line every 60 seconds. perform that into some file, have that
> file's contents processed and injected into MRTG, and there you are.
>
Indeed, I can highly recommend vmstat. It'll show almost anything you need
besides network load. If you have enough technical understanding, than read it's
man page - it's beautifully short (but good!) ;-) . vmstat is also very
comfortable for loggin your data into a file ready-to-be-parsed-later (vmstat
-n).
If you have a high sys% which you wanna check, one thing to do is to use tcpdump
to see perhaps faulty machines floding your server with errornous request (and
forcing your serever to reply something similar to "cant do that - error...").
Or just the most loading machines. I encountered this recently in my work.
Boaz.
=================================================================
To unsubscribe, send mail to [EMAIL PROTECTED] with
the word "unsubscribe" in the message body, e.g., run the command
echo unsubscribe | mail [EMAIL PROTECTED]
