Hoping for some guidance on a suggested architecture to try to avoid this scenario please:
Production room with primary SAN and storage (3 hosts) running production VMs.
Secondary DR room with capacity to store all production VMs and to run VMs required in a DR situation (single host).
Both rooms are on the same L2 subnet and are a mile or so apart connected by 10GbE.
We're currently licensed for vCenter Foundation and 6 sockets of Standard (but importantly the support for this is based on the initial purchase of a "HA Bundle" back in the days of 3.5).
I'm interested in using vSphere Replication to replicate from Room A to Room B but I cannot for the life of me figure out how I handle a scenario where I lose my vCenter server, which ideally I'd run in Room A since it's the production room.
People have suggested run vCenter from Room B, I've also had suggestions to go with something like an Essentials Plus bundle to cover the second site and to run with a vCenter in each site.
Could I get some input on the pros and cons of each approach please?
We've got a similar configuration here, though at 2 different physical sites. The best way I've found to recover so far is to manually copy over the .vmx files for the replicated VMs when configuring replication. This allows you to login to the DR ESXi host directly with the standalone vSphere client, add the VMs to inventory and start them up. I've tested this so far with Windows Active Directory servers and Exchange 2010, without any issues on startup.
Our vendor recommended backing up the vCenter Server database and manually re-installing the vCenter server at the DR site on failure of the primary site, but I haven't found a way to backup and restore the replication server and database, so it seems to me as re-installing the vCenter Server would have limited benefit in doing an actual restore of the replicated VMs.
I also tried replicating the vCenter Server appliance to the DR site, but on startup it said too many changes had been made and services wouldn't run or start correctly.
I agree with Joelrob action plan, just to add, if vmx file too is not available, just create new vm with no vmdk and add the replicated VMDK from DR site to this VM. This should allow to power on the VM successfully.
I have some doubts about this design and how will be recovery options that we will have in a disaster case.
About the Solution Design, should I have 2 "vSphere Replication Server", one per rooms? What will be the benefits with one more "vSphere Replication Server", as the appliance is a "vSphere Replication Server" also?
Should I keep "vSphere Replication Appliance" in the same "room" as vCenter?
About the recovery:
Lets imagine that I have running on my room1 ESX the vCenter and the "vSphere Replication Appliance". And I have on Room 2 the recovery room.
1) If I lost the room 1 were is located "vSphere Replication Appliance" and vCenter, I just need to add machines to the inventory of the room 2 host using .vmx files?
2) Is it recommended to have vCenter also replicated, so I can bring it back in the another room?
3) How should I proceed when the room 1 returns back?
I do not recommend using vSphere Replication to replicate vCenter Server in a single vCenter Server environment. While in some cases you may be able to "stitch together" the replicas of the .vmx and .vmdk files and make it work, this is not a VMware supported operation. In other words, if you do this and it fails, VMware Support may or may not be able to assist.
My recommendation is to use something like VDP (included with vSphere 5.1 Essentials Plus and higher) to perform a daily backup of the vCenter Server VM (Windows or Linux appliance). VDP has an "Emergency Restore" feature that enables restoring a VM directly to a host without vCenter Server being online. I realize the RPO with this type of approach is 24 hours and the recovery time is probably an hour or two, i.e. it is not a perfect solution, but at least it is supported.
Here is your answer:
Basically, on the replicated side, rename the files, register the .VMX and then power on the VM, answer the question and you are good to go.