Hi Everyone,
I am after a bit of help on this one.
I have a cluster of several hosts, HP DL 580 G5s and G7s
The issue I have have is that in spite of DRS set to fully automated and an agressive migration threshold, the guest machines will just continue to sit on the G7s until they are alarming on memory utilisation. The other G5 hosts are not even 30% used in terms of ram and CPU barely registers.
Vmotion is working as I am able to vmotion guests around within the cluster without an issue.
What am I missing ?
Regards
Stuart
Under the Cluster config, is the Virtual Machine Options set to default(Fully Automated)?
They are indeed set to Default (Fully automated)
Any DRS Groups? Rules? EVC? Please explain your set up a little more.
Ok, 6 hosts, mixture of G5s (128 GB) and G7s (256 GB) All running latest firmware, and running 4.1 U2. Vcenter is 4.1.0
EVC Mode is Xeon Core 2
As i said, fully automated and agressive migration policies. This flows down and is reflected in the machine machgine options.
There are no rules, or affinity/anti affinity rules. This is a VDI setup if that makes any difference, with XP SP3 clients running.
Vmotion is enabled and works on every cluster member without issue. There are no alarms or alerts of any type.
I've seen this a few times. Give vCenter a restart and see if that gets things moving again.
have you clicked Run DRS on the top right of the screen to push it to DRS immediately
You are not missing anything - remember DRS is not to evenly balance VM load across the ESXi host but is ensure that the VMs are receiving sufficient resources - so if they are receiving sufficient resources you can even have a situation where your G7s might be at 90% utilization and the VMs will not migrate -
I have a cluster of several hosts, HP DL 580 G5s and G7s
The issue I have have is that in spite of DRS set to fully automated and an agressive migration threshold, the guest machines will just continue to sit on the G7s until they are alarming on memory utilisation. The other G5 hosts are not even 30% used in terms of ram and CPU barely registers.
The default memory warning is at 82%, but the host is not memory-constrained until you hit 94% so there's no point in migrating a guest from your G7 to a G5.
If you do not have a memory or CPU constraint, there's no benefit to moving a guest and in fact, your guests may slow down moving to the G5 hosts because they have slower CPUs and slower memory.
I've got a cluster with identical configurations for every host and right now see 3 hosts in the 30-31% memory range and 3 from 83-85%.
In my opionion, the memory warning should not kick until much later and I just recently changed mine to warn at 90% and alarm at 94%.