How to troubleshoot vmware simple check in zabbix 2.4.5

[jamesNJ]

Hello all,

I am trying to use the vmware vmware.hv.perfcounter simple check and can't seem to get it to work for querying data stores.

Is there some way to easily test zabbix server simple checks? At least for vmware.hv.perfcounter the zabbix_get program returns unsupported item key. Is there either a way (like zabbix_get) to directly query simple check keys like this, or perhaps some easy way to get zabbix to execute simple checks easily and quickly for debugging?

Does anyone have working examples of how vmware.hv.perfcounter works with performance counters that take an "instance"? For example, I am trying to get used disk space on a volume called "TRex_ESX_Vol_1" using the following:

vmware.hv.perfcounter[{$URL},{HOST.HOST},"disk/used[latest]","TRex_ESX_Vol_1"]

and it doesn't return anything at all and the log doesn't have anything on this trigger.

Thanks!

[wiper]

Zabbix supports only two performance entities that can be queried with custom performance counter keys - hypervisor (vmware.hv.perfcounter) and virtual machine (vmware.vm.perfcounter). You could get the datastore space used by the target virtual machine with vmware.vm.perfcounter[{$URL},{HOST.HOST},"disk/used[latest]"] key, but there are no way to get the space used by the datastore. That would require a new key vmware.datastore.perfcounter and currently there are no plans for it.

Also note that storage utilization counters are only available from vCenter Server (not directly from ESX). See https://www.vmware.com/support/devel..._counters.html (there is also description what instance values can be used with this counter).

Regarding instance value – it's counter specific and in most cases using empty string for aggregate values would be enough.

[jamesNJ]

Wiper thanks for your reply.

I was struggling to figure out how the vmware docs correlated to various calls I could make with perfcounter, but I think I see now what you are referring to.

I think it is odd that a datastore usage metric isn't already available ... seems like an obvious thing to worry about, especially on active hypervisors where people overcommit storage with thin provisioning ... perhaps my local issues are a bit more extreme than others

Do you have any suggestions on where to best find this sort of data? Right now I'm looking at vmware's vCLI for linux, and digging around the SNMP capability to see what else is available.

Thanks!

[wiper]

Also during vmware service initialization the available performance counters are written in log (with debug level 4). Look for '.. adding performance counter <group>/<counter>[<roolup type>]' records. Although you would need to check the VMware documentation to find out if the counter is supported by hypervisor or virtual machine.

But except Zabbix (and manual SOAP requests during testing) I don't have any experience with VMware monitoring, sorry.

[NOB]

Hi

did you get any results from your investigations ?

We are in the same situation and would like to have the datastore
usage monitored as well.
I did the same thing like you (prior to reading your post) with the
same result, of course.

What about the experience with SOAP testing ?
I did use SOAP years ago and would have to dig into it, again.
Any help is appreciated.

Kind regards
Norbert.

[wiper]

I poked around a bit and found out that datastore counters are not 'real-time' counters and they must be handled a bit differently (basically the way how to get their refresh time, which is used by zabbix to query performance statistics).

So it's not as simple as adding just a new performance entity type. It would be easier to get datastore capacity/free space from configuration data. There is a prototype patch adding vmware.hv.datastore.size[<url>,<hypervisor uuid>,<datastore name>, <mode>] key in ZBXNEXT-2184, but hasn't been tested much.