load question

[eugenesc]

Hi Guys, im new to the forum but not new to VMware.

My name is Eugene Schenck I am a VCP and I work as a VMware Specialist at a company in Johannesburg, South-Africa.

I have a question around memory load on a esxi5 host. Well it’s more of a debate we have at work.

Scenario;

You have esxi5 hosts A and B each the exact same spec and with equal amount of virtual servers running. Both hosts run consistently 70% at memory threshold, these hosts share a vCenter server but live in different datacenters.

If host A fails in a DR scenario and business systems points their application load to the virtual servers on host B. What would happen, to host B memory utilization?

[Troy_Clavell]

if both hosts are running at 70% then if a single host failed the other host would have to pick up the workload of the first, therefore you being at 140% used.  If you are running the high of utilization you your hosts you may want to get a 3rd Host to help with your HA events.

[sparrowangelste]

Troy is right, you woudlnt be able to handle the load.

you should set the restart priority of critical servers to high in HA , or set some of the non-critcal servers to a lower vm restart.

[rickardnobel]

sparrowangelstechnology wrote:

Troy is right, you woudlnt be able to handle the load.

you should set the restart priority of critical servers to high in HA , or set some of the non-critcal servers to a lower vm restart.

In this case I do not think that will help. From my understanding the restart priority will only have a real effect for VMs with reservations. For all other VMs they will all be started, however the "high" some second before the others, but all will start and then the memory reclaiming techniques will have to kick in to handle the memory overcommit.

[eugenesc]

I understand the HA part etc, but was trying to hint towards the balloon memory driver and if the host would be over utilized?

Just so I understand this right, the 70% load I see in vcenter is the amont of configured memory as well as utilization from virtual servers? (what if the virtuals have been over specd and not using all of the given memory)

Sorry for the vague questions... I just want to get a better understanding if I claim that a host would cope with the extra load.

thx

[Troy_Clavell]

What is the actual resource utilization of the Host?  You can get a general idea by looking at the summary tab.  Also, highlight the cluster and selct the hosts tab.  You should get and idea of % memory being used.

balooning and host swapping would more than likely take place.  If that happens, in my opinion, it's not a good thing.  It means you have not properly sized your enviornment.

[eugenesc]

Ok the host in the above instance the memory is between 65 and 70%.

Also keep in mind that this is in an extreme as ive mentioned a DR senario.

neither of the two hosts are in a cluster or have HA etc. they are stand alone physicals connected to a vcenter.(for this example)

Thanks for the help thus far, I think I get the point - no physical can cope with a additional load exceeding its own 100% correct?

[rickardnobel]

eugenesc wrote:

Thanks for the help thus far, I think I get the point - no physical can cope with a additional load exceeding its own 100% correct?

That will depend somewhat on what you mean. You could certainly overcommit your physical RAM and the different memory handling techniques will do their best to keep everything going the best as possible.

So depending on the very specific needs of your VMs it might in fact run ok even with 140% memory load, however not recommended.

[Troy_Clavell]

no physical can cope with a additional load exceeding its own 100% correct?

resources can be oversubscribed.  However, this is all within reason, and depends.  With that said, with the scenario you put forth, I don't see anything good coming out of it.