VMware Cloud Community
JulietDeltaGolf
Enthusiast
Enthusiast

When does DRS beat HA

Hi,

We are currently implementing a stretched storage system where we will end up with say 5 ESXi hosts in site 1 and 5 hosts in site 2 and the storage will be available across both sites. The recommendation is to use the DRS host affinity rules to keep VMs that are primarily on site 1 storage running on hosts that are on site 1 and only failover to site 2 when site 1 is down. However we are trying to work out what might happen in the event of loosing some but all hosts on a site.

As a worse case example imagine that 4 of the 5 hosts running in site 2 go offline. What we imagine will happen is that HA will try to restart these VMs and will look at the host affinity rules for these and see that they have 'should run on host' set and then will try and start them all on the 1 remaining host in site 2, then once they are up and running DRS will kick in, seeing that single host is overloaded and shift some of the workload over onto site 1 hosts.

Does that sound right? Is that the behaviour that we would see in the real world?

Cheers.

Reply
0 Kudos
1 Reply
JulietDeltaGolf
Enthusiast
Enthusiast

Also we'd like to try and understand at what point does DRS start ignoring the 'VM should run on host X' rules - I.e. will it start VMs on a host until it is overloaded or will it start VMs on a 'non-preferred' host after the preferred one is X% utilised?

Reply
0 Kudos