Migrate VMs running in old FC-SAN to new FC-SAN via storage replication

[souvik1979]

All,

We are having one migration in pipeline. Where we have to migrate production application VMs from existing FC-SAN(NetApp,IBM) to new FC-SAN (Hitachi).

Current Infrastructure:- around 1000 of VMs are running on following hardware.

Hardware:- IBM HS21/22 BladeCenter, IBM rackmount server x3650M3,x3850M5

ESXi Version:- 4/4.1

vCenter Version:-4

VM Guest OS: Windows 2000,Windows 2003, Windows 2008, Linux 5.7,Linux 6.1

FC-SAN:- NetApp,IBM

New Infrastructure:-

Hardware:- HP BL460c Gen8 Blade

ESXi Version:- 5.1

vCenter Version:-5.1

VM Guest OS: As-IS

FC-SAN:- Hitachi VSP

High Level Approach:-

We'll built & make ESXi servers ready to host all of these VMs from old hardware & old FC-SAN (IBM ,NetApp)

1 Collate Migration Information

2 Connect VSPs to Legacy Arrays

3 Present Universal volumes to VSP

4 Discover External Volumes in VSP

5 Create and allocate Volumes to Host PiT copies.

6 Prepare & Test Hitachi Shadow Image Horcm configuration files

7 Migration Window Opens

8 Check Integrity of existing environment - Shut down hosts

9 Create ‘quick’ Replicas

10 Mount and test PiT copies

11 Run Remediation Procedure

12 Migration Window Closes

13 Delete Pairs and Disconnect External Volumes.

14 Continue to the Next Planned Volume Group

15 Decommission Legacy Storage

Once this storage activity completes then replicated LUNs will be visible to new ESXi 5.1.

What issue /steps needs to be performed to make these VMs up in new infrastructure.

[f10]

Are you are doing a block level copy of old luns ? If yes, then the first step would be assign an unique id to each VMFS volume by doing a resignature. You will get this option while mounting the VMFS datastores on Hitachi San.

You would then have to register all the VMs from the VMFS datastore on Hitachi San. There are many powercli scripts available to register VMs. However consider the following:

VMs with RDMs

VMs with VMDKs from more than one VMFS datastore.

VM mapping to specific resource pool.

Any VM used for Linux/Windows MSCS clustering.

-f10

http://highoncloud.blogspot.in

About VMware Virtualization on NetApp

[souvik1979]

Thanks f10 !

To answer to your query " Are you doing block level copy of old luns?" --Yes

So first we have to do resignature of those replicated datastore ? these will not bring any downtime to datastore right? any other precaution are required?

I'm in the process of capturing the full inventory of existing infra. So not sure about following.

VMs with RDM

VMs with VMDK more than one datastore

VM mapping to specific resource pool

Any VM used for Linux/Windows MS cluster

What exactly need to be checked for these above mentioned cases?

[souvik1979]

Anyone is having any further update on this discussion?

[taylorb]

Seems like you are doing this the hard way.

Easy way is to connect old and new storage to some VM host and do a storage vmotion.    You can connect just a few hosts and use them as "gateways" and vmotion VMs onto the host that can see all LUNs and then storage vmotion from LUNs on old SAN to new SAN.  It's a little bit of manual work, but less risky that what you are looking to do. 

[souvik1979]

Hi Taylorb,

Thanks!

But this is not possible in our case. Source storage is managed by other vendor & they are not ready to present storage directly our ESXi host. In order to mitigate we have opted this route. Sync & replicate storage at the backend ,then allocate LUNs from our storage system.