Basic
Assumptions:
The
customer does not necessarily need access to historical performance or event
data and is willing to sacrifice that.
The
customer is willing to accept minimal downtime so long as it is
planned.
1. Backup
the entire environment, including the VMs and the supporting systems and
databases. (!)
2. Stand-up
the new hosts with either 5.5 or 6.0
3. Stand-up
new datastore storage for your new 5.5 or 6.0 cluster.
4. Designate
one of your new hosts to be the transition host or ‘landing zone’
5. Add
an FC HBA to this landing zone host and have it zoned so that it can see the
existing VMFS3 datastores.
DO
NOT UPGRADE VMFS if prompted or offered!
6. Select
a number of non-essential virtual machines to serve as a
proof-of-concept.
7. Take
note of which datastore(s) the identified virtual machines reside.
8. Systematically
schedule the shutdown of the identified virtual machines.
9. Once
the virtual machines are powered-off, right-click and remove from
inventory.
DO
NOT DELETE. Remove from inventory.
10. On
the landing zone or transition host, browse the datastore where the VM to be
migrated resides, open the folder and find the configuration (.vmx) file.
Right-click on that file and choose Add to Inventory.
11. Once
the VM shows up in the new cluster, attempt to power it on. Verify that the
power-on works and the system is available on the customer’s network. Note that
the network port-group labels and such may be different between the old cluster
and new, so you might have it edit the VM’s settings to ensure the correct
port-group(s) are selected.
DO
NOT UPGRADE VIRTUAL HARDWARE OR VMWARE TOOLS AT THIS TIME.
12. Repeat
as necessary until all virtual machines are moved to the new cluster.
13. Plan
an upgrade of the VMware tools (requires a reboot) on each virtual
machine.
14. Plan
an upgrade of the VM virtual hardware level (requires a second reboot) on each
virtual machine.
15. Utilize
VMware’s Storage vMotion to move all of the VMs to the new
datastores.
16. Remove
the legacy VMFS3 datastores.
17. Shutdown
and decommission the old hardware.
I
have done this before with 5.5 and assume that it would operate the same way
with 6.0, but that is another risk that would need to be identified with going
right to 6.x. You could upgrade to 5.5 and then, once completed, upgrade to
6.0.
Note
that if any VM has an RDM, that will need to be handled separately. You can use
the same process, but before you are able to decommission the old storage you
will need to either migrate the external RDM to a new virtual VMDK (create new
VMDK, use guest OS tools to move the data) or another form of storage based on
the new array’s capabilities.
No comments:
Post a Comment