Cheshire Cat Computing

Hello I'm seeing something kind of strange with the check_vm.pl plugin. I'm using it to monitor two ESX4 hosts and a VCenter server in Nagios. I have several services defined for each: alarm status, datastore usage, memory and CPU. Originally I was getting the 'Perf stats not available at required interval' error when trying to collect CPU and memory reports for each host, but after reading through these forums I figured out the problem was the logging level on my VCenter server. When I changed the statistics for the 5 min interval to level 4 the check began returning both CPU and memory on both my hosts.

Now a couple days later I am seeing 'Perf stats not available' again...but only for the CPU check on 1 of my ESX hosts? CPU for the other host is still fine, as is memory for both. These checks are single service definitions applied to both hosts, with --server=$HOSTADDRESS$ added to the command definition to allow them to share a common config file.

What could have made one of my hosts stop collecting/retaining 5 min Perf stats?

Check the connection between this ESX host and the virtual centre. Is it sending logs? Use your virtual centre to try and view the CPU graph for this ESX server, can it view the graph for live data? If not, then it means the virtual centre really doesnt have the data which is why the plugin cannot obtain it.

The 'cannot retreive stats' message is what you get if the call to the API for the data retrieves nothing. This can happen if the logging level is too low (as you have already seen), or if it is not retrieving data at the 5min interval at all, or if the server is for some reason not sending data (maybe it has lost contact with the virtualcentre) and so on.

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

Yes, I can view Real Time logs from both esx hosts in VCenter for both CPU and memory (I just tried this for memory while the mem check is showing the Perf Stats error, logs come up fine). It certainly looks like my servers are sending the logs.

For clarification where is the best place to monitor CPU and Memory usage? Should I run the plugin targeting the vcenter server or the hosts directly? The Nagios XI wizard says for CPU/Mem you should target the ESX hosts but other checks should target VCenter, is this correct?

The way the plugin is written it should always be targetted at the VirtualCentre. The hosts might give slightly more up to date infor on mem and cpu, and of course can report it with the VC being down or unavailable, but they have a lot of other things missing. Although you can point the plugin at the host and it will work for some things, you need to be careful which you pick.

Can you successfully get any stats at all? Possibly the user you are authenticating as does not have sufficient rights to retrieve the stats, or the authentication is failing?

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