Cpu usage on W2K servers

[joyeux666]

Hi,

I'm a newbie and i would like to create a CPU usage graph for windopws 2000.

For my linux servers, i use ucd/net cpu usage and it runs correctly

I wanted use this template for my windows 2000 servers but i recuperate none data.

I search on the forum without success

First question, Can i use this template for cpu usage on W2K servers ??

What OID do i recuperate for this graph ??

Thanks for your help

sorry for my english

regards

David
France

cacti server
cacti 0.8.6.g
red hat 3 enterprise

[BSOD2600]

Windows does not use net/ucd snmp. Add the 'SNMP - Get Processor Information' associated data query.

[ezoltan]

David,

Probably the easiest way to do it is this:

1. On the Windows box, download and install SNMP Informant Standard from http://www.snmp-informant.com. This is a very good *free* SNMP agent. (btw, you must have SNMP installed first *before* you install the agent). You can use SNMP Informant Advanced if you need to monitor more then the standard counters (i.e. how many terminal server users are currently logged in), but it is not free.

2. In Cacti create a device for your Windows box (which you have probably done already)

3. Create a data source as follows:
-use "SNMP - Generic OID Template" as your data template
-for your Data Source Maximum Value leave the default of 100, as %CPU time will never go beyond it
-for OID use .1.3.6.1.4.1.9600.1.1.5.1.5.1.48 for the first processor, .1.3.6.1.4.1.9600.1.1.5.1.5.1.49 for the second, etc., or .1.3.6.1.4.1.9600.1.1.5.1.5.6.95.84.111.116.97.108 for the overall average

4. Create the graph using the "SNMP - Generic OID Template"

5. Add it to your graph tree and let it run for at least 2 polling intervals, then you should see your CPU time graphed.

You can use the above procedure for virtually any counter available via SNMP.

You can use Getif, a handy little *free* tool to test SNMP (and many other network-related things), before you create your graphs. I use it to test whether SNMP works on the monitored host, as well as to graph live data (i.e. poll the host every second). It can be downloaded from http://www.wtcs.org/snmp4tpc/getif.htm.

NB: If you choose to buy the SNMP Informant Advanced edition, the OID's will be different. You can use Getif to get it right.

Hope this helps.
Zoltan

[ezoltan]

David,

One more thing. In Windows, the counters referred to in my previous post will tell you how busy the CPU is as a percentage. It is an integer between 0 and 100, where 100 means the CPU runs flat out. No conversions or formulas needed, just graph the value as GAUGE.

For a more thorough monitoring you will need the advanced counters, which include CPU Queue Length (how many instructions have been queued for execution - should never go beyond 2 for an extended time), as well as Disk Queue Length (how many disk access requests have been queued to be processed by the storage subsystem - should also stay under 2). These two counters will show whether you need a faster CPU or disk/disk controller. Only looking at %CPU Time is not enough to find the bottleneck in a slow machine. But it comes with a price - have to buy the Advanced agent.

Microsoft gives some guidelines as to what to watch and what are common thresholds to look for. Where to find it:

1. Click Start -> Run, type PERFMON <Enter>. This brings up Performance Monitor.

2. Click Help -> Help Topics. This brings up the Performance Monitor Help.

3. Navigate to Performance Logs and Alerts -> Concepts -> Using Performance Logs and Alerts -> Setting up a monitoring configuration.

For threshold guidelines:

1. Navigate to Performance Logs and Alerts -> Concepts -> Using Performance Logs and Alerts -> Choosing Which Counters to Use

2. Click on the entry of interest (e.g. Finding processor bottlenecks)

3. In the right pane click Analysing Performance to view recommended thresholds.

All these counters are available via SNMP with the Advanced agent, and can be graphed in Cacti.

Have a great monitoring.

Zoltan