Automated System Usage Report Generation

[Gamoji Raghavendra]

Hi guys,

I am SA for 9 HP-UX 11.00, L2000 Servers each with 8 X 18.2 gig disks, 3 GB RAM and PA-RISC processor and kernel is 64-bit. Now I want to generate an automated report on usage of these servers on a weekly basis.

Below is the data which I need to collect :
1. CPU Usage (min, max and avg over the week)
2. Memory Usage (min, max and avg over the week)
3. Disk Usage (min, max and avg over the week)
4. Network Usage (min, max and avg over the week)
5. No of users logging in and out (daily)
6. No of unsuccessful logins for root user with IP address of the client (daily)
7. No of successful logins for root user.

I know I can use glance for 1 to 4, but I donot know how to use glance scripts. So it would be great if somebody can help me/point me to the right resource on this topic.

Coming to 5-7 I want an efficient way of handling this.

Any help in this regard would be highly appreciated.

Tnx,
Raghu.

[Sanjay_6]

Hi,

Suggest you should look at enablingthe audit on your system.

Hope this helps.

Regds

[Gamoji Raghavendra]

Hi Sunjay,

Thanks for the advice.

I can enable auditing on the system, but I read somewhere that auditing hinders the performance of the system.

Performance is a critical issue on all these servers. That is why I cannot audit enable the system.

What all I am intending to do is through simple shell scripts.

tnx,
Raghu.

[Sanjay_6]

Hi Raghu,

Whatever script you use, it has to poll the system to collect the type of data you are looking for and that is going to eat into the system resources. How much of the resource it is going to use depends on the polling interval and the code you use in your script.

Though i doubt this type of data extraction would be easy. What you are aiming for look quite high to me. Is there any particular reason you are looking for this type of performance data.

Regds

[Michael Tully]

Hi,

5. use the 'last' program
6. 'lastb'
7. Have a look at the /var/adm/sulog
/var/adm/syslog/syslog.log and 'last'
I would suggest that you disable root
logins except from then console.

For 1-4 I suggest you at least look at
measureware (glance plus) which provides
exactly this type of performance
information. You can install a trial
copy from your application CD set
without doing a re-boot. You could
look at using 'sar' but depending on
the type of hardware your using can
actually provide meaningless information.

sar -M -u (for multiple cpus)
sar -b (buffer)
sar -d (disk)

Using glance + you can specify your own
reports for management. I use it quite a
lot and this also helps with capacity planning.

-Michael

[Gamoji Raghavendra]

Hi Sanjay,
what you told was absolutely right. But the frequency at which I want to extract the data is very less, probably once/twice in a day.
So I donot think it would eat much of the resources.

Anyhow, I will study more about auditing.
Thanks for ur advice.

Micheal,
Thanks to you too ;)