Memory ballooning

[tobyjosham]

We're running a 5 host cluster, 6.5 U2 and the amount of memory available seems fine however a number of vm's are being adversely affected by memory ballooning.

If there is enough memory available why is it using ballooning?

[a_p_]

There could be several reasons for this, e.g. that a VM has a memory limit configured in its settings.

André

[tobyjosham]

There's no limits set on the vms.

If there's 100GB available on the host why would it resort to ballooning?

[a_p_ · Accepted solution]

With sufficient free physical memory, and no memory limits configured in the VM's settings (or on a resource pool if applicable) you shouldn't actually see ballooning.

Have these VMs previously been vMotioned from another host, which may haven been low on memory?

André

---------------------------------------------------------------------------------------------------------

Was it helpful? Let us know by completing this short survey here.

[tobyjosham]

A new host was recently added to the cluster and prior to that the available memory was short. Since then DRS has moved the vm's around and everything seems fine.

The cluster has 1.87TB of which 1.25TB is used and 1.75TB is assigned to the vm's. Surely this is enough without causing memory ballooning

[a_p_]

AFAIK the balloon driver will not actively release ballooned memory. It will deflate on request, i.e. when the guest OS requires more memory, and there's sufficient free memory on the host. Migrating (vMotion/DRS) will not help. The easiest way to release the ballooned memory is to reboot the VM (if this is possible in production).

Anyway, since the guest doesn't seem to need the ballooned memory for other processes, you shoulbn't see any issues, and may delay a reboot until the next patch day.

André

[tobyjosham]

Your previous message reminded of something.

Approx 1/2 of our vm's are in 3 project resource pools which have limits set.

The developers have done what developers always do and spun up too many virtual machines so even though the hosts and the cluster itself has enough, the resource pools are limited causing the ballooning and swapping.