New Nutanix Block AHV

  • 23 August 2016
  • 6 replies
  • 1142 views

Badge +8
Hi All,

We are about to secure our first Nutanix Deal (in our region) - our existing customer have a vsphere environment with 35 Virtual Machines (Exchange / AD / SQL / Oracle DB) - they are willing to go AHV.

After setting up the first block - what would be the best way to migrate all virtual machines from existing vspehre cluster?

Existing vSphere cluster is using VMFS data stores off their existing SAN. will be grateful for your suggestions / insights / best approach towards this.

Kind regards

This topic has been closed for comments

6 replies

Userlevel 7
Badge +25
I would be talking with your Account team for the specifics since it is a commercial opportunity, but I moved workloads for CE by just mounting a Container via NFS and moving the VMDK (or an svMotion of the whole VM).

I don't know if you could use parts of the "Cluster convert" workflow to then change those VMX/VMDKs into AHV guests if you prestaged NGT/VirtIO.
Userlevel 6
Badge +29
Folks - Moving this from Community Edition to the Commercial forums.

Congratulations on the upcoming purchase!

There's a couple different ways to slice this. For Exchange and Active Directory, those are apps that you generally dont ever "change platforms" with those applications, instead, you rebuild new instances and move the services over. This is the same from P2V or VMW to Hyper-V or Xen Server to VMW, or whatever. If the sides are different, then its best to rebuild.

For the other applications, you can either manually move each of them, look at "migration" keyword search on portal.nutanix.com for our KB articles on moving Windows to AHV, OR you could use a tool like Sureline Migrator, which works GREAT, and is the only tool out there that has AHV integration.

Happy to keep the conversation going as you plan this deployment.

Cheers,
Jon
Badge +8
Hi Jon - Many thanks for your reply and support.

I have a design / sizing related question:

For one of my prospective (SMB) customer with 3 x ESX Hosts and 65 Virtual Machines (Exchange / SQL / Oracle - JDedwards and others) -- I have come up with the following config:

Exchange = 400 MBX (version 2010)
Oracle JDedwards = used company wide with around 200 users maximum
SQL - is used by an internal / in-house built ERP (.Net Application + SQL Server)
all other virtual machines are test / dev + a virtualised TMG for content filtering.

2RU - 4Nodes (NX-1465-G5)

8 x Intel Xeon Processor 2.1GHz 8-core Broadwell E5-2620 v4 20M Cache32 x 16GB DDR4 Memory Module8 x 6TB 3.5" HDD4 x 800GB 3.5" SSD4 x 10GbE Dual Base-T Network Adapter
with the above config, i get 128GB Ram per node and approx. 19TiB overall usable capacity with RF2.

My Questions:

a) Are NX Nodes as reliable a hardware platform as Dell / Lenovo Nodes?
b) 6TB Drives - Incase of a single drive failure - rebuild time idea approx.?
c) I have selected 4th Node in order to provide some room for Node-failure plust additional compute for their expansion.
d) Shall I create separate container for MS Exchange and enable inline-compression?
e) What shall I size the CVM as per node? go with the defaults or ?

If you require further information - I can provide - Many thanks for your valuable feedback / support!

Kind regards
vM

PS : I am converting this customer to AHV + VNX.
Userlevel 6
Badge +29
Before I answer your question, just curious, how did you actually get to that sizing? Did you use sizer.nutanix.com? or some manual method. Again, not saying good or bad, just curious.


a) Are NX Nodes as reliable a hardware platform as Dell / Lenovo Nodes?
From the data we see from call-home, Yes, and keep in mind, a lot of the components are common between all platforms, like Proc, Memory, HBA's, and Disks. Differences are around motherboard, fans, and power supplies.

b) 6TB Drives - Incase of a single drive failure - rebuild time idea approx.?
This is hard to say, since we do not use RAID. If that 6TB drive is only 1TB full, we donly need to rebuild 1TB of data, not entire 6TB

c) I have selected 4th Node in order to provide some room for Node-failure plust additional compute for their expansion.
Ok, good.

d) Shall I create separate container for MS Exchange and enable inline-compression?
Have you reviewed our Exchange on AHV best practices: http://www.nutanix.com/go/optimizing-a-microsoft-exchange-deployment.html

Check that out first, and you'll find it will answer your question. Its a good read, courtesy of 

e) What shall I size the CVM as per node? go with the defaults or ?
Generally, the defaults are going to be just fine, especially with the 1000 series.
Badge +8
Hi Jon - thank you for your replies and apologies for the delayed response.

Yes, I used nutanix sizer to get an idea, plus observed their current work loads (cpu / ram / disk) utiilsation via vcenter / vsphere client - discussed their peak times etc with their IT ops team. Got a list of their Virtual Machines and vcpus assigned to get an idea of how they are utilising their existing specs.

Capacity: calculated the capacity + factored in their growth (didn't include storage efficiency when sizing capacity).


With Nutanix Sizer, it has 3 types of VM (Small / Medium / Large) - is there a way to input the exact vcpu's assigned for a current virtualised environment? for example either input via csv or manual (where we input virtual machines + vcpu + ram) ?

What other sizing tools would you recommend for future physical or virtual workloads?

Best regards
Userlevel 6
Badge +29
RE List in Sizer
No, but they are improving sizer constantly, so it will come eventually.

RE Tools for sizing
There are a lot out there, Dell DPACK is very good. Some other storage vendors (when coming from non-Nutanix to Nutanix) have good outputs, like EMC's MiTrend reports.

If you dont have something like that where it can give you real time aggregates stats, you can go off of static configuration/stats with RVTools. I did an internal video for our SE's on how to leverage RVTools to aggregate requirements to be able to put them in sizer. You can work with your local SE to tag-team that, and if they are curious, the video is part of the "WOTS" series (episode 16 I believe), so they can track that down.