vSphere 4 clustering question

[adamhodgkins83]

Hi all,

I'm currently working on a vSphere 4.1 environment that needs a re-arrange of the clusters (2 clusters, each with 16 nodes).  I've just come into this environment to find that there are a few blade enclosures with all of the cluster primary hosts sitting in one enclosure.... not ideal to say the least!!

I need to find a way of re-arranging the cluster with minimum downtime and effort involved, so far I have come up with 3 different proposals for the work but wanted to ask around to see if anyone has done this before, how they did it etc...

Option 1 - Move all of the VM's around to allow the removal of all hosts from the clusters and rebuild completely.  This I know would work but I can see it being very time consuming.

Option 2 - Use the das.preferrefPrimaries command - the major thing for me on this one is that I don't think that it is currently supported by VMware.

Option 3 - Try migrating all VM's off 4 out of the 5 primary nodes then put them into maintenance mode, letting the cluster re-elect new primary nodes.  With this idea when the original primary nodes come back 'online' will the cluster re-elect them back or would it be easier to rebuild the nodes (given that this will take around 10 mins each...)

Any thoughs/suggestions would be greatfully welcomed!!!

Thanks

Adam

[Linjo]

Or upgrade to vSphere5 where the whole concept of Primary/Secondary is gone and replaced with a more robust function...

// Linjo

[adamhodgkins83]

Hi Linjo,

I had thought of that but at the moment it isn't an option so I'm stuck with working with 4.1

Thanks

Adam

[kjb007]

Option 3 would be the easiest to start with, but maintaining the primaries in 4.1 with nodes distributed this way is cumbersome.  The re-election would occur as those original primaries would no longer be up and running, and they should not go back to the original ones once the role is taken over by another node.

Option 2 is unsupported, but would accomplish your task in an easier manner.

Option 1 would be similar to option 3.

-KjB

[farkasharry]

I would create 4 clusters with 8 nodes. Take 4 nodes from each blade chassis for a 8 nodes cluster.
With this, you can ensure that you always have a 5. primary in another chassis!
Downtime or speed is depending on the amount of VM's running of those hosts. I am sure that vMotioning around the VM's a bit, will allow you to split up and create new clusters by not effecting the VMs uptime. Of course, only if all hosts are seeing the same LUN's and networks...