Optimal clustering considerations

[ASchinkel]

At present I have 5 ESX hosts configured into 2 clusters of 3 and 2 hosts respectively.

The hosts are running ESX v4.1 and vCenter is v4.1 as well.

My concern is this is a waste of resource potential with the additional overhead and limitations it puts on both clusters.

The clusters were initially set up like this so that we would have one cluster to run a test upgrade on before upgrading the other.

What I am considering is to consolidate all 5 hosts into one cluster.

To meet the needs of being able to run a test installation of future upgrades:

- remove one host from the cluster of 5 and upgrade it, migrate a VM to the upgraded host and ensure it works

- break full redundancy in the cluster and remove a second host, add it to a cluster with the first, and test again

- if the first two tests are a success, upgrade the remaining three hosts, and add the two test hosts back into the primary cluster

I see no immediate problems with running things in the above configuration, and would appreciate any insight to additional considerations I may want to take into account.

Thank you,

-Aaron

[FranckRookie]

Hi Aaron,

I can't see any problem with what you propose. Just take care not to overload your cluster to be able to remove one or two hosts for testing.

Good luck.

Regards

Franck

[ASchinkel]

I have 3 BL460c G1 half height blades and 2 BL480c G1 full height blades.

Can the failover contraints recognize between 'small host with 20gB ram' and 'big host with 36gB' ?

[FranckRookie]

Yes. When an HA event occurs, every host knows which VMs were on the failed ESX. They check if they have enough resources to restart the machines depending on reservations and admission control rules.

Franck

[AndreTheGiant]

Two small clusters (with 2 and 3 nodes) is not so good...

Five is a good number and enough also to implement DPM (if you have the license).

Andre