I wanted to get some feedback from other users regarding the display of the Memory Utilization metrics for VMWare VM's.

After moving to vmware host monitoring in opmanager we noticed that we were getting conflicting information, specifically relating to the Memory Utilization monitoring. (The same applies to CPU Utilization monitoring also, however the difference is not as pronounced).

After much consternation, (and discussions with manageengine support) it turns out that the memory utilization details in the Virtualization>All Devices list view, as well as the Memory Utilization Dial in the vm snapshot page, is using the performance stats provided by the Host, while the alerts and monitors were using SNMP/WMI from the virtual machines to report memory utilization problems.

In the case of memory utilization this is using the Active Memory statistic from VMware. According to the vmware documentation, this Active Memory refers to the memory pages that a virtual machine has "touched" within a recent time period.

As an example of the discrepancies that this causes, our Exchange 2010 CAS server has 16GB of memory and consistently runs at around 73% usage. However the Memory Utilization dial only shows 15% usage. The discrepancy is due to Exchange grabbing a chunk of memory ready for use, (73%), but only a portion of these memory pages are considered active by vmware, hence the 15% usage.

I question whether using the vmware active memory stat is relevant in the list view, and in the vm snapshot page for a number of reasons.

Firstly the vmware active memory statistic documentation refers to being the number of memory pages that have been "touched" recently. However there is no indication of the time scale.

Secondly, while this is an interesting stat from a vmware host or vCenter level, it is fairly irrelevant at the individual virtual machine snapshot level, where ideally the information presented is internal to the virtual machine in question, not from the host.

Thirdly, having the alerts and monitors use SNMP, but the list view and snapshot dials use vmware stats, is confusing and frankly counter-intuitive.

So I would be interested to here what other users have to say on the issue, and whether it would be worth asking ManageEngine to change this behaviour. There are several options for the change.

Firstly it would be nice to have the option to choose which stat could be displayed in the list view or snapshot dials. Whether this is a global options, template based, or at the individual vm.

Secondly the VMWare Active Memory stat could be displayed in a seperate dial, for individual vm snapshots.

Thirdly this stat could be displayed at the host or vCenter level.

 

I believe this is what another forum user is referring too, under post "Memory Snapshot Bug". Cheers