I received a question a while back about DRS initiated VMotions. One of my customers wanted to know which metrics were used by DRS for deciding if a VM needs to be VMotioned to a different host or not. These metrics are:
Host CPU: Active (includes run and ready Mhz)
Host Memory: Active
Just a little something that’s nice to know I guess. I need to dive into the actual algorithm that is being used by DRS and if I can find some decent info and have some spare time on my hands I will definitely write an article about it.
I’ve been curious about this also so thank you. I’m suprised that such an important metric like CPU Ready is not used. Is there documentation on what the stars are equal to for migration threshold?
By default, every 12 minutes (5 times per hour) DRS makes an evaluation. Good article Duncan. Short and to the point!
“This message will self destruct in 5 seconds”
I’ve think I’ve read somewhere that VMware is working on adding I/O (maybe both networking and storage) to DRS metrics in the future. Can anyone confirm this?
http://www.yellow-bricks.com/2009/09/16/io-drs-providing-performance-isolation-to-vms-in-shared-storage-environments-ta3461/
@JArenth,
i agree, my farm was running well from a Vsphere client point of view, when the users complain.
The issue was than DRS recommandations with a more than 3 stars [vmware recommandation], was not moving any VMs, even the CPU ready was higher than 2000ms on almost all VMs on 2 of the 4 hosts.
Amazing that DRS has nothing to deal with CPU ready, when this is the first counter to look at in the Performance Troublesshooting for VMware vsphere 4
http://www.vmware.com/resources/techresources/10066
Changing to the most aggreesiv behavior, the VMs was moving around every 30 minutes !
Hi.. I want to the threshold value, how it is calculated and what is the max values for deviation, threshold etc….
Thanks..