Missing Counters in vROps

[TheVMinator]

Under virtual disk in vROps I only see read latency (ms) and write latency (ms).  I would like to see all the counters that are available in ESXTOP (KAVG, GAVG and DAVG)

https://kb.vmware.com/selfservice/microsites/search.do?language=en_US&cmd=displayKC&externalId=10082...

Is it possible to enable these extra counters?

Thanks!

[dtaliafe]

You should have these disk metrics on the host objects in vROps.  If you don't see them make sure they are enabled in the policy.

KAVG - Disk | Aggregate of all instances | Kernel Disk Command Latency (ms)

DAVG - Disk | <device> | Physical Device Command Latency (ms)

GAVG - not sure if there is one that maps directly to this, but it is just the sum of KAVG + DAVG

I've asked VMware support and our TAM about this many times.  I think they need better documentation for many of the vROps metrics and an easier way to find how metrics in vROps map to counters in esxtop or vCenter.

[TheVMinator]

Great info - thanks.

In light of this article:

http://www.vmignite.com/2016/02/vrops-6-how-to-enable-hidden-metrics-and-properties/

Is it possible there are "hidden" metrics that pertain to the "virtual disk" object that are not enabled by default?  Ideally, I'd like to see for each virtual machine, what the latency is when it sends commands to storage, and what portion of that latency is in the host, and what portion is below the host.  I'd like to see this not just for the entire ESXi host but for each VMs storage traffic individually.

Is it possible this information could be part of a "hidden" metric?

[dtaliafe]

There are a lot of metrics disabled by default.  You can enable them in your policy, but I don't think these are exactly what you're looking for.

The Virtual Disk | Total latency and Read/Write latency will give you the latency for each VM.  The latency at the host level and outside of the host isn't measured at a VM level as far as I know.  The kernel average (KAVG) is going to be the same for each VM on the host since they all share the same kernel.  Likewise the device average (DAVG) will be the same for each VM sharing storage on that physical device.  So I think you have to look at the host level for the latency outside of the VM.

[TheVMinator]

OK thanks again.  So if KAVG and DAVG are the same for each VM, then GAVG is also the same for each VM, and so wouldn't read_latency be the same for each VM?  If read_latency is different for each VM, then how is vROps calculating read_latency?  Using vmware tools inside the OS?