I'm doing a greenfield deployment of VSAN and I have a bit of a challenging situation. Initially we're going to start out with a 3 host cluster. 2 of the 3 hosts we're wanting to use long-term are currently bare metal installs and hosting important applications, the other is an R730xd purchased just for this purpose. I see 2 options:
Any suggestions or possibilities I've missed?
I would go with option 2. If I understand, you get vSAN running on the new host with vCenter on it and FTT = 0. You then P2V the bare metal to the vSAN with FTT = 0. Then comes the high wire act of formatting the bare metal to ESXi and adding them to the vSAN cluster and the changing the FTT = 1. The danger comes if you have a failure as you're adding the two nodes into the vSAN cluster.
A little twist on this, what I did. We have 2 SSD and 14 disks in each host. I created a RAID 5 using 3 of disks and made that a local data store. That's where I deployed vCenter and where you could then P2V your bare metal. Then I did vSAN on the rest of the disks and added my other hosts. That way you can at least mitigate a disk failure taking you down. Eventually I storage vMotion the VMs to vSAN, deleted the local datastore and claimed the disks for vSAN.
If that is you're only option, I guess you have to. Obviously not ideal. Thank you, Zach.
I would go with option 2. If I understand, you get vSAN running on the new host with vCenter on it and FTT = 0. You then P2V the bare metal to the vSAN with FTT = 0. Then comes the high wire act of formatting the bare metal to ESXi and adding them to the vSAN cluster and the changing the FTT = 1. The danger comes if you have a failure as you're adding the two nodes into the vSAN cluster.
A little twist on this, what I did. We have 2 SSD and 14 disks in each host. I created a RAID 5 using 3 of disks and made that a local data store. That's where I deployed vCenter and where you could then P2V your bare metal. Then I did vSAN on the rest of the disks and added my other hosts. That way you can at least mitigate a disk failure taking you down. Eventually I storage vMotion the VMs to vSAN, deleted the local datastore and claimed the disks for vSAN.
If that is you're only option, I guess you have to. Obviously not ideal. Thank you, Zach.
I hadn't thought of the scenario you mentioned and using storage vMotion - I like that route better so I'll probably do that. Thanks for the input!