Unix Cacti Server, Windows 2003 agents

[shaaad]

Are there any templates for windows 2003 servers? I'm having a hell of a time getting my windows 2003 machines to work with my cacti server. I cannot seem to get any data, and I'm not sure how I need to build my graphs or data templates to get it all working.

Any help much appreciated. Thanks!

Nick

[melchandra]

1. Have you installed and started the snmp service on the 2003 servers?
2. Have you configured it with a password?
3. Have you selected all the checkboxes on the "Agent" tab (properties of the snmp service)
4. Can you verify the snmp service is running and working with a tool like snmpwalk or GetIF?

[DevilSun]

Yeah, what he said. Basically make sure you can get SNMP info from them...you should be able to "walk" the OID of 1 to get the basic system info for the machine. Otherwise, you're having basic SNMP problems.

However, one thing I noticed was some of the security scripts / group policy stuff Microsoft has suggested for hardening servers (especially domain controllers) blocks traffic to most ports unless it's secure using IPSec.

First and foremost, get in there with SNMP like melchandra said.

[melchandra]

Actually, Microsoft has enabled a sort of firewall on the snmp service. On the properties of the snmp service, on the Security tab, there is an option to limit the hosts that can contact the server. I'm not sure if what you specify here is implemented via ipsec rules, it's a good guess, but may not be correct. Whatever the case is, it's a good security tool. We have all of our servers on a seperate VLAN, so we just use vlan filters to make sure that nobody else can get to our servers other than what is needed.

[Burke]

In addition to the SNMP stuff, you'll want to get Nagios compiled on your Linux machine and install nsclient on your Windows 2003 servers. Once you have the check_nt script available, you'll be able to create scripts that use it during your polling intervals. Do a search here for nsclient and check_nt to get more details and sample code from me and others.

[Guest]

The NSClient does make life easier, however, I can't get acurate results from it. Something like "Processor(_Total)\% Processor Time" always returns 100 and someothers return only 0. Some of the exchange perf counters are working great though. Any idea why the processor ones seem broken?

[DonPetry]

I am also experiencing a similar issue. In 8 Processor machines, NSClient is returing a false CPU Utilization of 100% using the Windows Performance Counter. The default check I've been using is a "Total" utilization of all processors:

\\Processor(_Total)\\% Processor Time

This returns 100%. When I run the check against an indivual processor in an 8 Proc machine, I still get 100%.

\\Processor(0)\\% Processor Time = 100
\\Processor(1)\\% Processor Time = 100
\\Processor(2)\\% Processor Time = 100
\\Processor(3)\\% Processor Time = 100
\\Processor(4)\\% Processor Time = 100
\\Processor(5)\\% Processor Time = 100
\\Processor(6)\\% Processor Time = 100
\\Processor(7)\\% Processor Time = 100
\\Processor(8)\\% Processor Time = 0

When I check beyond the total processors in the machine I receive 0%. Since the PerfMon on machine is registering normal CPU Utilization levels, this looks like a bug in NSClient's handling of our 8 Proc machines.

These checks are working fine on 1, 2 and 4 processor machines. Too bad NSClient is no longer supported... I can't wait to move to a nice SNMP stack.