DRS power-on algorithm?

[rrosenkoetter]

So, I'm in the last couple of weeks of migrating my last farm to VI3 from ESX 2.5.x...

I migrated about 30 more VMs this weekend to a cluster of 9 existing host servers. I have DRS set for it to decide the best place to power on VMs...

As I powered on the VMs (this was over an entire day of migrating), I noticed that DRS kept chosing one of my host servers a lot more than the other host servers...

When the migration was done, I noticed that one host server showing yellow in Virtual Center... I went and looked at all the hosts...

Host 1 - 15 VMs

Host 2 - 15 VMs

Host 3 - 20 VMs

Host 4 - 21 VMs

Host 5 - 16 VMs

Host 6 - 18 VMs

Host 7 - 15 VMs

Host 8 - 29 VMs[/b]

Host 9 - 14 VMs

Hosts 1-7,9 all had %CPU in the 26%-44% range, Host 8 was at 64%

Hosts 1-7,9 all had %memory in the 48%-72% range, Host 8 was at 82%

So why did DRS continue to power on VMs on Host 8 instead of using the other host servers?

I manually migrated some VMs off of Host 8 to distribute the load better... but I wish I knew how this worked and why it distrubuted the load that way.

[oreeh]

Do you have any DRS rules set?

[rrosenkoetter]

Yeah, I have a bunch of rules for the VMs that were already on those host servers... but nothing that should have applied to the 30 VMs I migrated over this weekend.

[oreeh]

Do any of these rules contain more than two VMs (there's a bug with those rules)?

Try removing the rules or change them to only reference two VMs - this should start the automatical migration of VMs to other hosts.

[rrosenkoetter]

I knew about the bug with trying to keep [b]separate[/b][/i] more than 2 VMs at a time... so I got rid of all those rules a while back...

I do still have rules that keep [b]together[/b][/i] more than 2 VMs though. They actually seem to be working... Is there a bug with those too?

[oreeh]

I'm not sure if this bug affects rules that keep VMs together, but... I personally (just to be safe ) only use rules (regardless of the rule type) containing two VMs.