DRS does not work

[alexandru_ilie]

Hi,

             I have configured a cluster with 4 vSphere 5 Enterprise hosts,with DRS and HA, but it seems that DRS does not kick in. There are many VMs on one host, so DRS should kick in and start migrating the vms, but nothing happens. I have tried manually vMotion between the hosts and it works fine.

When i created the cluster, no DRS and HA was activated. I have done this after adding the hosts, but i do not see this as an impediment. The DRS is set to aggressive.

             Could someone give me a tip? Where could i look to see the actually problem? All configuration seems to be ok.

             vMotion and the Management Network are on the same vSwitch. The vSwitch is using 2 physical nics. The nics are configured as TEAM for vMotion and Management Network, Nic1 being the main nic for Management Network and Nic2 in standby, and vice versa for vMotion.

Thank you.

[alexandru_ilie]

Could it be because the load is not the high on the host holding the most of the vms? The loads are 29%, 11%, 6% and 2% when it comes to memory. And when it comes to cpu it is bellow 5% on all hosts right now. However, i expected DRS to do something with that 29% memory on the first host, to start migrating vms on the other hosts.

[a_p_]

There are many VMs on one host, so DRS should kick in ...

Not necessarily. DRS is not meant for load balancing, but for resolving resource contention or satisfying DRS rules.

André

[weinstein5]

As Andre indicate DRS will only kick in if the VMs are not receiving enough resources - so theoretically if you have 40 VMs in a 3 node DRS cluster all 40 VMs could be running on  a single node of the cluster if the all VMs were receiving the necessary resources.

[BenLoveday]

Hi there,

Andre hit the nail on the head when he said that DRS is not a true load balancer.

It will only apply rules if it deems that there is a resource contention issue and moves VM's appropriately to improve performance. To be honest I was expecting the same thing the first time I enabled DRS :-)

The only way to help better spread your VM workload would be to create custom rules that either use VM affinity or anti-affinity, this way you could keep VM's together such as an app server and db server that talk to each other, or you might say that a group of heavy IO vm's can't all reside on the same host.

Cheers,

Ben Loveday

[jdptechnc]

DRS will only kick in if there is enough of an imbalance to justify doing so.  If you put one of your hosts in maintenance mode, DRS will kick in at that time to migrate your VM's, that should be a good enough test.  If you manually move all of your VMs on one host, maybe that will be enough of an imbalance to watch an automated DRS take place.

[alexandru_ilie]

Thank you all for all your answers.

Best regards, Alex. :smileywink: