VMware Cloud Community
shinacalypse
Contributor
Contributor

Migration from ESX 4.0 cluster to ESXi 5.1

Hi!

I'm very new to VMware products, though I've done a bit of training and bought some books. Now I have a project dropped suddenly in my lap.. to move an existing ESX 4.0 based cluster (5 hosts, and NetApp storage) to shiny new ESXi 5.1 hosts (3 of them). Two of the new servers are supposed to be at the primary site, and the 3rd in a failover site. The 5.1 hosts are to use the same data stores as the 4.0 servers.

There are about 45 virtual machines in the old cluster.

I'm wondering what is the best way to do this?

Im thinking I should

1. upgrade the 4.0 VCenter VM to 5.1,

2. add the new 5.1 ESXi hosts to the current 4.0 cluster,

3. migrate the vms to the new ESXi hosts (vmotion, storage vmotion?),

4. remove old hosts from the cluster,

5. upgrade vmware tools on the VMs.

The problems I have are the actual migrations, and how to setup the 3rd esxi host as disaster recovery. Is any of this possible like i described it? What should I watch out for, and what are the pitfalls?

Any assistance would be appreciated!

Thanks!

Reply
0 Kudos
4 Replies
fr0nk
Contributor
Contributor

Hi,

Take a backup of the vCenter 4.0 DB (no kidding). Then ensure your DB is supported in 5.1: http://pubs.vmware.com/vsphere-51/index.jsp?topic=%2Fcom.vmware.vsphere.upgrade.doc%2FGUID-93612C3C-...

If your DB isn't supported anymore do an export and import to a supported version. When your vCenter DB isn't running on the vCenter Server you just have to adjust your 64(!)-Bit DSN.

Don't use the automatic install (which installs SSO, vCenter Inventory and vCenter in that order).

  • Start the SSO Setup.
  • DON'T CHANGE THE TABLE NAMES, just execute the script delivered on the install ISO. You may change the DB name.
  • DON'T use: "The following characters are not supported in passwords: semicolon (;),  double quotation mark ("), single quotation mark ('), circumflex (^),  and backslash (\). Passwords must comply with Windows Group Policy  Object (GPO) password policy.". Link: http://pubs.vmware.com/vsphere-51/index.jsp?topic=%2Fcom.vmware.vsphere.install.doc%2FGUID-200B9E03-...
  • The Passwords of your RSA_Admin and RSA_User users obviously must comply with your (MS)-SQL password policies.

Here's a catchya: If the lookup of your AD fails during SSO install, simply add it afterwards. Don't delay that task.

Then install the Inventory Service, you can't do anything wrong with this one.

Then upgrade-install vCenter 5.1 using the DSN if your DB is remote.

Install the Web Client.


Install the new vSphere Client.

Install or upgrade your existing Update Manager installation and database

Depending on the hosts you have: Look for a custom vendor ISO that may include special VIBs for your hardware (10Gbit NICs, etc).

  • Import the ISO to UM. Create a baseline in this process.
  • Attach the baseline to the Cluster or ESX hosts.
  • Scan.
  • Remediate one host at at time. Test with one hosts, especially for missing drivers (e.g. you're doing iSCSI on a 10Gbit NIC that hasn't driver support built-in).

vMotion between Clusters and 4.x and 5.x hosts is fully supported. Use it.

svMotion is supported between VMFS-3 and VMFS-5 datastores. Use it.

Don't upgrade your datastores to VMFS-5 until you evicted or upgraded the last 4.x node from the cluster. 4.x hosts can't read VMFS-5. 5.x hosts can read VMFS-3 and VMFS-5. The Blocksize (1-8MB) of VMFS-3 datastores remains the same, even if upgraded. VMFS-3 partitions upgraded to VMFS-5 remain MBR until they breach the 2TB size. Then they become GPT. If you used other block sizes than 1MB in your VMFS-3 datastores consider wiping and re-formatting them with VMFS-5. That's the only way you get to 1MB blocksize.

VMware Tools upgrade to 5.1 doesn't force a reboot if VMware tools are present on the guest.

Consider upgrading your VM hardware to Version 9. Check if there are virtual peripherals or configurations that aren't supported anymore (e.g.: VLANCE vNICs).

Normally, no downtime is required at all. during this process when admission control was turned on for your old cluster.

Hope this helps,

Kind regards,

Frank

Reply
0 Kudos
shinacalypse
Contributor
Contributor

Wow.

Thank you, thank you, thank you for the detailed analysis, Frank. Before I went on to irreparably harm something!

I'm still going through what you said, but I have to ask, is the way I want to do it the best possible way?

Is there an easier and safer, way (even if more time consuming)?

Thanks a lot!

Also I don't fully understand how to do this:

"

Depending  on the hosts you have: Look for a custom vendor ISO that may include  special VIBs for your hardware (10Gbit NICs, etc).

  • Import the ISO to UM. Create a baseline in this process.
  • Attach the baseline to the Cluster or ESX hosts.
  • Scan.
  • Remediate  one host at at time. Test with one hosts, especially for missing  drivers (e.g. you're doing iSCSI on a 10Gbit NIC that hasn't driver  support built-in).

"

Can you please give more detail?

Reply
0 Kudos
fr0nk
Contributor
Contributor

Easy doesn't always go well with safe 😉 So in short: no.

Depending  on the hosts you have: Look for a custom vendor ISO that  may include  special VIBs for your hardware (10Gbit NICs, etc).

  • Import the ISO to UM. Create a baseline in this process.
  • Attach the baseline to the Cluster or ESX hosts.
  • Scan.
  • Remediate   one host at at time. Test with one hosts, especially for missing   drivers (e.g. you're doing iSCSI on a 10Gbit NIC that hasn't driver   support built-in).

I guess you have no idea how to use the update manager. This should give you a pretty good walk-through: http://www.youtube.com/watch?v=ds7fvpGT1Sg

If you have any questions in detail, just ask.

Cheers,

Frank

Reply
0 Kudos
shinacalypse
Contributor
Contributor

Hi Frank

and thanks again for responding. I have really not much of an idea of what I'm supposed to do and this has been made a performance requirement for me, so I'm kind of petrified. Smiley Happy

I just need some reassurance that what I plan to do is the best possible way of doing it, with zero disruption to the existing system.

Thanks for the youtube link!

I'm supposed to use 3 servers to consolidate the 5 original hosts, and turn one into a disaster recovery server.

Please how is this done?

When I add the esxi 5.1 hosts to the esx 4.0 cluster, do I still have to setup networking and datastore access or does configuration get passed automatically?

I apologize for the silly questions, and I'm frantically reading all I can online.

Thanks again for listening!

Reply
0 Kudos