
We have 8 Dell PowerEdge 2950 servers. Some of them are for our forthcoming RM CC4 network, and some have active roles in the current network. Two of them were purchased before the rest, and use 3.5” hard drives rather than the newer 2.5” disks that the rest have. All servers were bought with 2x 70GB disks and configured for RAID1.
We wish to make the most out of the servers and retire some old hardware, and VMWare ESXi is the way we are going to do it.
The roles for the ones involved in our plan are as follows:
A: BBS-ESX-001 (Currently the only VMWare server; 2.5” 10k SAS)
B: BBS-TSS-001 (Terminal Services server for remote access; 3.5” 15k SAS)
C: BBS-ZCS-001 (Email server running Zimbra; 3.5” 15k SAS)
D: BBS-SVR-001 (RM CC4 Forest Root; 2.5” 10k SAS)
Servers B and C were the original servers; and they have the same disk type. These will become VMWare hosts. In the event of hardware failure, in theory we can just take disks from the failed server, stick them in the working one, and things are back up and running in a relatively short space of time.
(If we had a SAN that could support all our VMs we’d probably use it. Testing our Hitachi SMS100 iSCSI SAN to serve VM images as well as running the main RM CC4 network is something for a later date.)
To maintain the use of everything we currently have and end up with two VMWare hosts with 300GB of capacity on each, we devised a cunning plan. Probably not as cunning as a fox who used to be Professor of Cunning at Oxford university; but a plan none the less.
- Install 2x 300GB drives into (A)
- Virtualise (B) onto (A) using P2V
- Install 2x 300GB 3.5” drives into (B)
- Install and configure VMWare ESXi onto (B) - name it BBS-ESX-002
- Move all VMs from (A) onto (B)
- Migrate ZCS to run physically on (A) on 300GB volume
- Install 2x 300GB 3.5” drives into (C)
- Install and configure VMWare ESXi onto (C) - name it BBS-ESX-001
- Balance load by moving some VMs from (B) onto (C)
- Install 2x 300GB drives into (D)
Some stages in the plan must be done when there is low network utilisation, and others can be done during peak times as it is unlikely to affect anything at all.
I will also take this opportunity to upgrade the server operating system for the Zimbra email server from Ubuntu 6.06 to 8.04.
