"host available memory" alarms - how critical?

[kacsp]

I've recently installed Veeam ONE and am using it to monitor our ESXi Hosts (no vCenter). The Hosts aren't that heavily used, but are perhaps overcommitted in terms of allocated memory, and Veeam ONE is now alerting on one Host with an "host available memory" Alarm.

So I'm trying to relate this "host available memory" Alarm to exactly how serious this is in terms of running low on Host memory resources, and the knock-on affect this will have on our VMs. In terms of figures, in Veeam ONE, 'Memory Active' is showing at around 8% whilst 'Memory Usage' is around 90%. From what I can see the "host available memory" Alarm relates to the 'Memory Usage' metric, but my understanding is that Memory Active is the more critical metric.

So if the 'Memory Active' figure remains where it is at a fairly low level, is there anything to really worry about with this Host, even though it's 'Memory Usage' is high?

Thanks

[Shestakov]

Hello kacsp,

You are correct, "Host Available Memory" alarm is purely based on "Memory Usage" counter. Memory Active shows a sum of all active memory metrics for all powered-on VMs + vSphere services. Do you have any not VMware-related services running on the ESXi?

I would also take a look at the 2 more metrics:
Memory Management: Memory Balloon if >0, host is out of memory and the hypervisor tries to handle it slowing down VMs` guest OSes.
Memory Swap Rate is used as a last resort to reclaim memory from VMs due to the limitations on performance. If > 0, definitely need to take action.

Thanks!

[kacsp]

Hi Nikita, I'm not sure what exactly you mean here:

Do you have any not VMware-related services running on the ESXi?

'Memory Active' on the Host is showing at around 8%, according to Veeam ONE. That Host runs a number of VMs plus ESXi overhead, which will all contribute to the 'Memory Active' total.

But my main question is, if the 'Memory Active' figure remains where it is at a fairly low level, is there anything to really worry about with this Host, even though it's 'Memory Usage' is high? What I mean by that is will it cause me a problem & negatively impact my VMs if 'Memory Usage' remains at 90%, or even rises towards 100%?

And if 'Memory Usage' can safely be left to run that high without any problems, can I just ignore or disable the "host available memory" Alarms altogether?

Thanks

[Shestakov]

What I mean by that is will it cause me a problem & negatively impact my VMs if 'Memory Usage' remains at 90%, or even rises towards 100%?

The answer is: "Yes it will", since the Memory Usage is showing the amount of used memory on the host. If there is no memory, VMs` tasks will not be executed. You should warn if the amount is 90%, since the metric is average for the last 5 minutes. Probably in some shorter instances there were 100% pikes. Check which VMs consume the most of memory.

As I mentioned above, Memory Active is a sum of all active memory metrics for all powered-on VMs + vSphere services. And this is an estimation made by VMkernel based on number of recently touched memory pages. Note that Memory Active is originally measured not in "percents", but in "GB". If you still have a big difference between Memory Active and Memory Usage, probably some host memory is used by 3rd party services(that was my question in the 1st post of the topic).

Thanks!

[kacsp]

Thanks that's been helpful...and my apologies for mixing up the Memory Active unit. I've done some comparisons in Infrastructure View and Business View, and gotten a better understanding of my situation. Active is low, but Usage is high, and this is due to a number of VMs showing high consumed memory. So we do somehow have to reduce consumed memory of VMs on that Host.

I do have the option of cutting in half allocated memory to a couple of VMs though, which may give some breathing space. Can you suggest anything else?


My initial confusion over the Active/Usage situation partly stemmed from reading this article:

http://www.yellow-bricks.com/2011/04/29 ... ng-memory/

Regards...

[Shestakov]

The article is short and a bit outdated, but still fair. However, it states a point regarding Memory Active vs. Memory Consumed, while we were discussing Memory Active and Memory Usage.

In terms of figures, in Veeam ONE, 'Memory Active' is showing at around 8% whilst 'Memory Usage' is around 90%.

And once again, please note, that the Measurement Unit for Memory Active is Bytes, not percents. So, I`d double-check the value, probably Memory Active is not that low.

I do have the option of cutting in half allocated memory to a couple of VMs though, which may give some breathing space. Can you suggest anything else?

As I said, for better understanding/ I would also take a look at the 2 more metrics:
Memory Management: Memory Balloon if >0, host is out of memory and the hypervisor tries to handle it slowing down VMs` guest OSes.
Memory Swap Rate is used as a last resort to reclaim memory from VMs due to the limitations on performance. If > 0, definitely need to take action.

The actions to do include: Decreasing allocated VMs memory, planing the load distributing tasks along the whole day, adding more physical memory to the host.

Thanks!

[kacsp]

...it states a point regarding Memory Active vs. Memory Consumed, while we were discussing Memory Active and Memory Usage.

In Veeam, 'Memory Consumed' is expressed in GB whilst 'Memory Usage' is a %, but aren't they referring to the same thing, that being the amount of memory that's been used up on the Host? So if I can reduce my 'Memory Consumed', I will in effect also be reducing 'Memory Usage' at the same time?

Sorry to labour the point here, but I'm just trying to understand exactly what Veeam ONE's telling me.

I've no Host issues with Memory Management: Memory Balloon or Memory Swap Rate at the moment, but nevertheless, as 'Memory Usage' is around 90%, I will be taking action to reduce consumed memory in the ways you suggest. Thanks again...

[Shestakov]

You are correct the metrics are alike, to be more precise on the Host-level Memory Usage is a Percentage of available machine memory = consumed / machine memory size, while host Consumed Memory = total host memory - free host memory. Thus reducing the memory usage will decrease both of them. Check it here.
Thanks!

[kacsp]

Thanks again Nikita...

[Shestakov]

You are welcome, I do understand where the confusion comes from.
Once you have a question, don`t hesitate asking!