Need some help guys - its realy important

[orauxpro]

Hello everybody

Can anybody Please tell me How to do the following issues in hp-ux900. Version 11.

1) Detect Operating System Problems
2) The key statistics to extract from any operating system monitor such as CPU load , device queues , network activity (queues) , memory management (paging/swapping) ,
3) how to Look at CPU utilization to see the ratio between time spent running in application mode and time spent running in operating system mode.
4) how to Look at runing queues to see how many processes are runable
5) how many system calls are being executed.
6) How can I See if paging or swapping is occurring.
7) how to check on the number of I/Os being performed.

Thank for your help, I realy appriciate it.

Oraux.

 

[unixqueen]

Hi orauxpro,

1) This would depend on the problem, /var/adm/syslog/syslog.log will contain some, as will dmesg. NETTL logs contain network related errors(you need to use netfmt to read it though). There are lots of other places

The rest of you questions are performance related, we could give you multiple commands to check things, but to get a reel feel for how a machine is performing you should collect statistics over a period of time.
I suggest that if you havn't got Measureware, Glance and Perfview installed, install the free 60 day trial versions on the quarterly support plus CD's. You could leave Measureware running for say a month and then extract the data to Excel format, load it into Excel and produce graphs.

Day to day running top will give you some info, as will Glance.

Regards
Queenie.

 

[MikeLacey]

Hi Abdull,

Your first question is a littlw wide I'm afraid - you detect problems on a
UNIX system by looking at many different things so there's not a simple
answer to your query.

The rest of your questions:

There are three UNIX commands that should help you here.

vmstat - process, virtual memory usage, swap, system vs user cpu utilization

iostat - disk i/o statistics

netstat - nework i/o statistics

Each of these commands should be run with the same option '-1' and they will
display a line, or lines with iostat, of output every second. For example
'vmstat -1' and 'netstat -1'

Please look at the man pages for each command for an explanation of the data
they prpduce. I would direct you particularly to the vmstat command which
has much useful information.

I hope that this information will be useful to you.
Mike
michael.j.lacey@ntlworld.com
Email welcome if you're in a hurry or something -- but post in tek-tips as well please, and I will post my reply here as well.

 

[Guest_imported]

Wow, not asking for much eh???? A lot depends on your h/w config and your applications. Multiple cpu, disks/lvm/array, running as a multipurpose system or a database server or?

Check your logs: syslog, btmp, utmp, wtmp for basic of what is going on. lastb -25. There are many, many, many possible things to check.

swapinfo -n tells you about your swapping.

I agree with the above, get glance, measureware. They are the best method to capture relevance information. It is probably installed, just find the binary glance or gpm (if you are running xwindows). The text version is not as pretty but I think you can see more data from all the different screens.

Top, netstat, swapinfo -n, bdf, etc etc etc (no pun intended) are just the begining. I run sar to capture data for analysis. Sar is a posting unto itself, so just read the man pages and play around with it. It takes a while to use but it is comprehensive as hades.

 

[orauxpro]

Thanks guys for big help, i realy appriciate it.