Best way to move VM from an old Blade environment to a NEW one

[vmware128]

Hello,

We are going to be moving to a new HPBLC7000 with ProLiant BL460c G7 blades. Currently We are running our environment on IBM Blade center H. The SAN will stay the same as it's fairly new. so Questions I have are regarding how to move the vm's off of IBM blade to HP blade. We are currently at esxi 4.1 and our vcenter server is a physical machine using sql server 2005 express. I would like to take this opportunity to upgrade to 5.1 and make the  VC a virtual machine with its' Database on a separate VM. So my questions are how do I accomplish this using a best practice approach. Some questions I have are:

1) do I need to even bother importing current VC DB onto new VC?

2) Do I need two virtual centers or Do I just add the new HP Host to current vcenter and vmotion(if CPU are compatible, if not what are my options?)

3) I'm just confused if I need to first move vm's onto new blade and then upgrade create new Vsphere on HP blade and then somehow migrate my vm's over?

I just need some guidance on how to best go about this. Thank You

[a_p_]

Assuming you don't need the old vCenter Server's data (history, events, resource pools settings, permissions, ...) and further assuming the hosts aren't compatible anyway, I'd go ahead and install the new hosts as well as the new vCenter Server and present the datastores to the new hosts. Then shutdown the VM's on the IBM hosts, detach them from the inventory, attach them to the new hosts' inventory and power them on. Make sure you select "I moved it" when asked whether moved or copied. Once done you may want to upgrade VMware Tools and the VM's hardware version (compatibility mode).

André

[vmware128]

Thanks Andre, So I'm assuming that there is no way to do this without having downtime because of the different hardware? Are there going to be any gotchas because the new vcenter and esxi hosts will be at 5.1 and the current vm's are running on hosts with 4.1?

[admin]

Donwtime will be required since they are different hardware and vMotion cannot be done.4.1 VMs will work fine on ESXi 5.1

[logiboy123]

Don't forget to turn on EVC in the newly built cluster before powering on VMs.

The above method is the quickest, easiest and most efficient.

Yes you will need downtime, that is unavoidable if you want to migrate correctly/safely.

[asherman]

Downtime won't necessarily be required for the VMs if he uses a compatible EVC mode on his HP blades. How old are the IBM blades? are they intel processors?

We just did this in our environment, going from HP BL460c G6 (ESX 4.0) blades to Cisco UCS B200 (ESXi 5.1) blades. In our case, we set the new UCS hosts to run in "Intel Nehalem Generation" EVC mode, and this allowed us to live migrate about 80 VMs from the HP hosts. Nobody even realized it happened, as the VMs will only drop a single ping, as with any other vmotion.

What will require downtime is the eventual vmtools and vm hardware upgrade that you'll want to do. It can wait for your next maintenance window, though.

I'd recommend building a new vCenter from scratch well in advance so you can take your time setting it up. Just use a trial license, it is good for 60 days. Once you are ready, you can convert your old vCenter 4 license to a vCenter 5 license and apply it to your new vCenter, replacing the trial key. Your old vCenter will continue to work, but you are technically unlicensed from that point on the old one.

We built our new vCenter as a VM on the new UCS blades. We ran into a "chicken & egg" issue setting the EVC mode since it is a cluster-level operation, and you need vCenter to do that. however, you cant change EVC level with powered-on VMs in that cluster! To get around that, we made a temporary cluster with a single host, moved new vcenter to it, made our change to the other cluster, and then moved it back.

Remember to configure your vswitch and datastores exactly the same in the new environment so you don't have problems during migration.

We used a kind of "life-boat" method to move the VMs from old environment to new. In the old environment, I disabled DRS and filled up a host with the VMs I wanted to migrate. Then, I disconnected the host from the old vCenter. In the new vCenter, I added that host into the EVC-moded cluster, and then migrated them off. They only dropped one ping. Once the life-boat host was empty, disconnect from new vCenter, reattach to old vCenter, and repeat until your old environment is empty.

[vmware128]

Few questions:

1)Currently our datastores are vmfs3, I will want to upgrade to vmfs5. What's the best procedure for doing this? If I create a new Datastore In the new vcenter, can I migrate using storage vmotion from vmfs3 to vmfs5?

2) The HP has Virtual connect which we intend to use so will this affect how the dvswitch needs to be configured on the new vcenter?Do IP addresses need to change at all during the migration if using virtual connect?

[admin]

Nothing needs to be changed on the dvswitch or network level. If you are using LACP, then please set the policy NIC Teaming policy to IP Hash on vSwitch and port group. Also once you have upgraded all your ESXi hosts to 5.1 and then have a look at this link http://www.vmware.com/files/pdf/techpaper/VMFS-5_Upgrade_Considerations.pdf

[asherman]

We left our datastores on vmfs3 after our migration, but now that we are a few weeks after it, we are using storage vmotion to move VMs onto new storage groups with new datastores. Once we move them all, then we will reclaim our old datastores and probably carve them differently. again, doing all of this during the day; no maintenance window required.

good luck with vconnect. we have been burned by it's buginess in the past, and learned that HP no longer has anyone that can answer the tough questions about it (they actually told us this!). All I can suggest is make sure your firmware is up-to-date before you put production servers on there.

[vmware128]

So what's the verdict then, downtime or as Asherman has suggested? The IBM are IBM eServer BladeCenter HS21 -[7995G6U] blades are Intel(R) Xeon(R) CPU           E5450  @ 3.00GHz

[admin]

if you have extra SAN to accommodate VMs, go what Asherman has suggested

[vmware128]

We don't have extra SAN Per say, we have room on the same SAN which we intend to attach to HP. I don't get this part ..Can you go into more details "We built our new vCenter as a VM on the new UCS blades. We ran into a "chicken & egg" issue setting the EVC mode since it is a cluster-level operation, and you need vCenter to do that. however, you cant change EVC level with powered-on VMs in that cluster! To get around that, we made a temporary cluster with a single host, moved new vcenter to it, made our change to the other cluster, and then moved it back." SO does the tenp cluster get turned on the old vcenter? or is the temp cluster on the new vcenter?

[asherman]

When you create your new vCenter, you are probably going to create it on the new 5.1 hosts in a cluster. If you then want to make a cluster-level change, you will be affecting the new vCenter VM, which you need to make the change in the first place. In the case of enabling EVC, we found that you can't do this with powered-on VMs.

To workaround it, we made a temporary cluster for vcenter to live in while we enabled EVC on the other cluster.

The temp cluster does not go in your old environment at all. This is just for your new environment.

[vmware128]

Ahh, so once EVC is enabled I move vcenter vm off of temp cluster and back to primary cluster. SO when choosing which EVC mode, do I need to choose what CPU's my IBM blades have or what HP has? IBM eServer BladeCenter HS21 -[7995G6U] blades are Intel(R) Xeon(R) CPU           E5450  @ 3.00GHz and my HP have Six-Core Intel Xeon, 2933 MHz. Thanks. Another question I have is , one of our servers which houses all of our users Home directories out on the SAN has like 7 virtual disks with a MAX size of 256 GIG. The Datastore is getting pretty full. When I move to the new HP, is there a better way to set this up? The 256 restriction is due to the block size so what's a good  practice for handling datastores becoming full? The disk on the Guest OS is a dymanic disk which we keep extended 256 gigs to as it starts becoming full.

[asherman · Accepted solution]

Yep, you've got it.

For EVC, you need to look at the CPU architecture that your IBM HS21's have. I think they are Penryn, but I could be mistaken.

On the HP blade cluster, you'll need to set them to use the Penryn (or whatever) EVC mode.  Once you have that set, try the lifeboat method on an IBM blade with a single test VM and see if you can live migrate it. If it doesnt work, you'll know you have to try another EVC mode.

I don't have much insight on storage, so I can't really speak to your second question.

[vmware128]

Thanks. You all have been a big help.