Cheshire Cat Computing

I have not yet been able to test against vshpere5 as we do not yet have it on our site.

The net stuff is not yet working as I've had no time to work on it. A rewrite is planned that will result in a separate agent which submits passive status notifications into Nagios, rather than an active check script, as the API performance is pitiful once you have a few hundred guests. At this point I should be able to make better use of things like network stats.

_________________
Steve Shipway
UNIX Systems, ITSS, University of Auckland, NZ
Woe unto them that rise up early in the morning... -- Isaiah 5:11

The command line looks OK and seems to be working - it is returning the statistic for the host. However, the NSCA support is only for GUEST data, not for host/cluster/datacenter data.

You can add the '--debug 1' option to get some progress info as it runs.

You can omit the --hostname option. In this case, you'll get info for ALL guests sent to the NSCA server.

You can look at your NSCA logs on the Nagios server to see what is coming in, and external command logs in the nagios log. You may need to use --canon, --tolower and --nscastrip to convert a FQDN to the guest hostname as use in Nagios.

I'm currently working on a new, completely passive agent to monitor VMWare via the API for Nagios and MRTG. This will give passive status info for all things (including hosts, clusters, datacenters) and will dynamically create config files if required. However it is some way off operation at the moment.

_________________
Steve Shipway
UNIX Systems, ITSS, University of Auckland, NZ
Woe unto them that rise up early in the morning... -- Isaiah 5:11

Pulling out network stats in a useable format is very awkward. Alternatives exist - you can monitor the network interfaces from the guest (either via SNMP or using something like nrpe or nsclient), or if you install the fully managed ESX Switch then the virtual switch acts just like a normal SNMP-managed switch, and you can use normal methods of monitoring as you would for physical switches.

Current development is around a new agent that interfaces directly with VirtualCentre, rrdcached (for MRTG) and livestatus (for Nagios). This generates configuration files and pushes stats and status directly in as passive alerts; this is more efficient, and can cope with over a thousand VMware objects at once. The development version of the new agent currently monitors cpu, memory, datastores, alarms and status, but not yet the awkward network and hba stats.

_________________
Steve Shipway
UNIX Systems, ITSS, University of Auckland, NZ
Woe unto them that rise up early in the morning... -- Isaiah 5:11

Hello,

I love your script but for some reasons It started to returning Unknown status.

We have vSphere 5.1.

Script worked fine for a long time. On 5.1 version too.

But now I keep getting this results:

./check_vmware.pl --mode nagios --config test-config.vmware --host my.hostname.com --report cpu

Return value: Perf stats not available at required interval (300s) or invalid instance

I tryed in debug mode:

[root@blabla /usr/local/icinga/libexec]# ./check_vmware.pl --mode nagios --config test-config.vmware --host my.hostname.com --report cpu --debug 1
Starting.
Connecting
Connected
Server Time : 2013-09-09T07:35:57.413242Z
Report type requested is [cpu]
Base is my.hostname.com
Retrieving PerfMgr data
Selected interval is: 300
Processing entities:
my.hostname.com
Creating query for my.hostname.com
Start time: 2013-09-09T07:36:00Z
End time : 2013-09-09T07:41:00Z
Retrieving data...
Disconnecting...
Exiting with status (3)
Perf stats not available at required interval (300s) or invalid instance.

What could be wrong?

Thank you and kind regards.

This is the result that comes when the plugin is unable to retrieve the perf stats from the Virtualcentre (if you point the script directly at an ESX server it is the same)

In the VC configuration, you can define which statistics are kept for hosts, and for how long. If I remember correctly, you can choose 'level1', 'level2' etc and say how long to keep 5min, 1 hour, 1day etc.

You should choose 'level2' or 'level3', and keep at least some 5min (smallest granularity) statistics, else the perfstat retrieval function cannot return any values. Some stats have 'current' values; the plugin goes for the 5min averages as it is likely to be getting called every 5min and so these will give the more meaningful information.

_________________
Steve Shipway
UNIX Systems, ITSS, University of Auckland, NZ
Woe unto them that rise up early in the morning... -- Isaiah 5:11