Lalegre
Virtuoso
Virtuoso

vSphere Stretch Cluster (No vSAN) - Upgrade

Jump to solution

I am in process of migrate from vSphere 5.1 to 6.0 on an Stretch Cluster. The customer is an ISP and for testing purpouses, it wants to upgrade first and entire site, wait one week and if there are no issues the second one.

Is there a possibility to have an issue on that week having a mixed Strech Cluster version? I mean having one in vSphere 5.1 and the other one in 6.0?

Thanks in advance

0 Kudos
1 Solution

Accepted Solutions

The important pieces to get right here are the interoperability between 5.1 and 6.0 at the vCenter and ESXi layer, in addition to any 3rd party products you're using. You essentially need to ensure you're always in a supported state.

First and foremost, ensure you always have vCenter and ESXi compatibility. The latest supported for your setup is 5.1 Update 3 of ESXi and 6.0 Update 2 of vCenter. Verify your 3rd party software, too.

In addition, someone mentioned vSphere HA differences between hosts at either site. Stretched clusters will only use the lowest common denominator with regards to feature sets, so there's no harm in running a mixed 5.1 / 6.0 cluster and their different feature sets, assuming you're sticking to the interop matrix. I had a problem with enabling APD and it was because there was  host in the cluster that didn't support it.

Other than that, as long as you stick to the interop matrix, you'll be fine with your plan.  Here's the product interop link if you don't already have it:

VMware Product Interoperability Matrices

----------------------------------------- Please consider marking this answer "correct" or "helpful" if you found it useful (you'll get points too). Mike Brown VMware, Cisco Data Center, and NetApp dude Sr. Systems Engineer michael.b.brown3@gmail.com Twitter: @VirtuallyMikeB Blog: http://VirtuallyMikeBrown.com LinkedIn: http://LinkedIn.com/in/michaelbbrown

View solution in original post

0 Kudos
5 Replies
daphnissov
Immortal
Immortal

I wouldn't rule it out, and generally I'd say for a stretched cluster going across such a leap as 5.1 -> 6.0 that's not a very good idea. If they wanted to wait on a separate cluster, that'd be one thing, but intra-cluster you want to have at the same major versions as soon as possible.

0 Kudos
Lalegre
Virtuoso
Virtuoso

Thanks for the fast response. Yes i thought the same as you did, but do you have any specific reason to don't do this?

I mean i say this because i can migrate the vms from one site to another one in different versions, but maybe there is something regard "Deep dive features"

0 Kudos
daphnissov
Immortal
Immortal

With such a radical change between those versions, HA and DRS behave differently in some respects, especially with the enhancements to how APD and PDL are processed. Plus, depending on your storage, you may have potential incompatibility issues on the "new" side versus the old. This isn't to say that you'll have a failure because of it, but you're increasing complexity and the time window in which such things could arise. Again, if they had a separate cluster which wasn't stretched that used the same storage and compute and they wished to test out the upgrade there, then that would be a lot better than doing it on half of a stretched cluster that uses vMSC.

0 Kudos

The important pieces to get right here are the interoperability between 5.1 and 6.0 at the vCenter and ESXi layer, in addition to any 3rd party products you're using. You essentially need to ensure you're always in a supported state.

First and foremost, ensure you always have vCenter and ESXi compatibility. The latest supported for your setup is 5.1 Update 3 of ESXi and 6.0 Update 2 of vCenter. Verify your 3rd party software, too.

In addition, someone mentioned vSphere HA differences between hosts at either site. Stretched clusters will only use the lowest common denominator with regards to feature sets, so there's no harm in running a mixed 5.1 / 6.0 cluster and their different feature sets, assuming you're sticking to the interop matrix. I had a problem with enabling APD and it was because there was  host in the cluster that didn't support it.

Other than that, as long as you stick to the interop matrix, you'll be fine with your plan.  Here's the product interop link if you don't already have it:

VMware Product Interoperability Matrices

----------------------------------------- Please consider marking this answer "correct" or "helpful" if you found it useful (you'll get points too). Mike Brown VMware, Cisco Data Center, and NetApp dude Sr. Systems Engineer michael.b.brown3@gmail.com Twitter: @VirtuallyMikeB Blog: http://VirtuallyMikeBrown.com LinkedIn: http://LinkedIn.com/in/michaelbbrown
0 Kudos
Lalegre
Virtuoso
Virtuoso

We already did some test, upgrading two nodes of Cisco UCS B-Series on one site using the latest drivers supported for 6.0 U2. The vCenter Server is currently in 6.0 U2.

We propose to do the upgrade of both sites at the same time. One node of Site A, one node of Site B and so on upon completion.

I will tell you about the results. Thanks!

0 Kudos