Looking at the documentation, it appears a direct upgrade from ESX 4 to ESXi 4.1 is not supported. So I was hoping to run my manual upgrade plan by the forum and see if it makes sense. I've got two ESX 4.0 hosts in a DRS cluster with plently of RAM. Vcenter is already at 4.1 and the rest of my ESX farm is on ESXi 4.1.
Here's the plan:
Sound OK, or would you suggest something different? The rest of my farm is on ESXi 4.1, so I don't want to upgrade to ESX 4.1 even though that would be easier.
Thanks for the feedback. So no known issues with ESX and ESXi combined in the same DRS cluster?
None that we have seen. We have mixed clusters now, and all is good. Remember, ESXi and ESX are the same hypervisor, so that should be able to co-exist in a cluster with no complaints.
Only if you are using FT mixing ESX with ESXi might be an issue as it expects similar build numbers etc. Other than that it is no problem at all.
Available now on Amazon: vSphere 4.1 HA and DRS technical deepdive
Could you elaborate on step 4 (unplug SAN and iSCSI ports to prevent accidental overwrites) . If i wipe out an esx 4 update 2 host and install esxi 4.1 how could a connection to san cause accidental overwrites?
Also for step 7, did you backup the host configurations when they were at esx 4 and apply them back to the hosts when they were esx 4.1 or what?
When you are installing ESXi, you could accidentally install the ESXi instance on the wrong LUN, which overwrites everything (the host will allow you to install ESXi on any LUN it can see). If you chose the LUN where all your VMs are, you just wiped out your VM datastore (hope you got a backup). I assume that you are not booting from the SAN? If you are planning on booting from the SAN, you will need your SAN connected when you install ESXi. I've found that an easier way to do this is to mask all the LUNs that are not the boot volume from the SAN itself (i.e. hide all the non boot LUNs so they are not even seen by the host). That way, you only have 1 LUN choice when you install ESXi with no risk of accidentally overwriting a VM datastore. Then after the installation, unmask the LUNs on the SAN.
Our hosts have a mirrored set of hard drives that have ESX 4 on, we do not boot to SAN. Our consultant setup NFS shares to use as datastores so i assume I should not be concerned about the accidental data loss?
What is your opinion of setting up the host configuration after the host has been erased and ESXi 4.1 installed, should i use host profiles or just document and rebuild the config manually?
If you are using NFS and not booting from SAN I don't believe you will have the problem of accidentally overwriting your VM storage, so you should be fine with that. As far as host profiles go, we don't use them in our environment (there was some sort of limitation they imposed that we decided we didn't like when we originally configured our ESX/VM environment, I wasn't involved in that so I don't have further details).
Raymond - it is my understanding the ESXi is the future of the platform, that eventually there won't be an ESX anymore. I'm not sure when this is going to happen, but we figured we wanted to get on board as soon as we can to avoid headaches later. Unfortunately this is causing some minor headaches now, though none of them are VMware's fault, mostly problems with our hardware vendor.
People are going to face this upgrade one day. It's certainly better to face it now if you have the resources, than having a major upgrade difficulty when for example, ESXi 5.0 comes out (and ESX 5.0 doesn't) and you suddenly need the functionality.
1) So making a host profile of a esx 4 host and then wiping it and loading esxi 4.1 and applying the host configuration should work?
2) The host profile is stored in the vcenter database?
3) I assume when i eventually replace the three current hosts with new hosts, i will not be able to use the host profile from the old host and apply it to the new host? I will be moving from IBM servers to Dell.
Host profiles are independant of hardware (dell to HP).
Anything that doesn't apply simply errors. For example, if I built a host profile on a server with four NICs, and applied it to a server with two NICs, it sets up those two NICs to reflect vmnic1 and vmnic2 on the first server, then errors about the others.
In general all these things will be fine.
What I plan to do is setup the 3 new Dell servers to run along side the 3 current IBM servers then cold migrate the vms to the new Dell hosts. We have 3 Virtual Switches one for VMkernel/Service Console (do i still need service console w/ ESXi?) one for a couple of vlans our VM's run in and one for iSCSI/NFS . Each virtual switch has two nics which act as primary/failover. Where i get confused about is how do i know what vmware will identify as vmnic0,1,2,3,4,5 on the new Dells compared to what is setup on the IBM's already? If i could connect the 6 nics to the correctly configured ports on our switch i should be able to apply the host profile as you mentioned exactly.