Hello!
ESXi 4.1 U2,
Cluster with 5 Hosts (3 R900/2 Proc/4 Cores/8 Logical/128 GB, 1 R910/2 Proc/6 Cores/24 Logical/128 GB, 1 R810/2 Proc/10 Cores/40 Logical/128 GB),
EVC Intel "Merom" Gen. (Xeon Core 2)
HA & DRS (Fully Automated @ priorities 1,2 & 3) on.
The problem is that the R910 and R810 are hosting so many VMs that the % Memory for those two hosts is VERY high. I haven't had any problems arise from this situation, but this can't be normal. What I wind up doing is manually migrating one or two VMs off those two hosts to the other Hosts.
Why isn't DRS balancing things out better?
Thanks!
Chris
DRS balances the VMs based on the resource needs. If the hosts are providing adequate resources to the VMs, then DRS has done it's job. If you wish to manually spread the VMs across the cluster, you can do so as well.
Keep in mind, If you are uncomfortable with having any particular VMs run together on the same host you can use the VM-to-VM anti-affinity rules.
I absolutely agree with the two of you. I just think DRS could itself divvy up the VMs better by making more use of the underutilized Hosts (the R900s) thereby relieving the other two Hosts.
I'm not used to getting the "The Host memory usage" warning in vSphere. This only started happening after I introduced the two newer servers.
I wonder if you bumped the migration threshold up to a 4 or 5 if this would help avoid the "host memory usage" warnings?