Guest memory alert and 100% reported active guest mem

[carlbeck]

I have a bunch of Win2012 VM's that receive "Virtual machine memory usage" Critical Alert. It seems that the alert appears when a new VM is added on the host they are located on

The machines have 16GB memory, with "reserve all memory" ticked, and are grouped in three on each host. The hosts are DL380g9 with 64GB installed memory.

I suppose the problem might be related to ESXi reporting 100% active guest memory, but why is this? If this was by design when "reserve all memory" is ticked, why generate an alarm?

I know that the way ESXi is measuring active guest memory isn't very accurate, but reporting 100% and generating an alarm sounds a little too much?

Thanks!

[unsichtbare]

Why ""reserve all memory""?

There are only a few use cases I am aware of that truly justify this setting. Moreover, if a host failed, HA might not be able to start VMs on a different host.

As long as there is no contention for memory (and there should be none with 3 VM's at 16GB RAM on a 64GB RAM ESXi host), reserving memory will not increase performance either. And since TPS is now disabled by default, there is no security vulnerability either.

[carlbeck]

Thanks for a quick reply,

Honestly I have no answer to why "reserve all memory", this is a request from the guys developing the applications that are running on those VM's. I'll check how they justify this, as you point out we have more than enough RAM on the host.

However, it can't be by design that guests with reserved memory generate "critical" alarms?

And what problems can the reserved memory cause HA? All hosts that are in HA clusters have capacity to solely run all VM's in that cluster.

[unsichtbare]

App guys always want all the memory! Their perspective is protecting the performance of their application, at the expense of the performance and availability of the environment!

You have to tell them NO, because this setting may preclude HA from functioning correctly. Moreover the use case for reserving all the memory on a VM is protecting one/several VMs among many. When reserving all the memory for the majority of VMs, it becomes self-defeating. Protect your infrastructure and your mission-critical apps will run better as a result!

[genoIE]

I have several clients with the same issue. They all have HyperFlex deployed and the "Storage Controller VM's" have 72GB of ram all of it reserved. They do not fail over as there is one on each host and they start and stop with the host. They are all showing memory usage alerts all of the time! They say it is annoying to constantly have alerts and I agree!