New cluster configuration questions

[symen]

I've been tasked with designing a  virtualization cluster.  I'm currently only passingly familiar with  vSphere; I will be sending someone to a training course, but the course  doesn't occur until after I need to have the design finished.  I think  I'll be okay, I learn fast, and I understand the underlying concepts,  but I want to try to ensure that I'm buying the right stuff for  everything to work before I spend the money.

My goals are to  build a platform to support 30 to 50 VMs, running a mixture of Windows  and Unix, for services such as Active Directory, file, print, web,  email, and various applications.  Redundancy for high availability is  desired as well.

We have two datacenters in two  different buildings, with a redundant 1GB (I think) link between them.   One has redundant electrical feeds.  I had thought to install a SAN  (Dell EqualLogic PS4000) in each location, synchronized with SAN  replication, and then four ESX servers (two in each location).  I want  to use Dell R510 servers with 2x quad-core Xeons and 48GB of RAM for  these.  I would use all four ESX servers all the time, but in the event  of a problem at either site, the unaffected site would become our DR  site.  I would like to have this failover happen automatically.

So, my questions are:

1. Is this even possible?  I have a book on vSphere, and based on what I've read, it looks like it is.

2. Will I need any other hardware beyond 4 Dell R510s, 2 EqualLogic PS4000s, and 4 good managed switches for the iSCSI traffic, with a dedicated fiber link between them, two large UPSs, and a few dozen CAT5 cables?  I already have a server to use for vCenter.

3. UPSs.  Is there a preferred brand that works well with vSphere?  I usually use APC units, but I don't have strong feelings about them one way or the other.

4. Which VMWare software will I need?  I'm a little hazy on what all the different packages do.  I'm guessing I need ESX Enterprise, and vSphere, but I want to make sure I'm not missing anything.

5. What is the best backup method for this?  Is SAN replication sufficient?  Or, should I use an external tape unit?

Thanks in advance for any help you can provide.

[bulletprooffool]

1. Is this even possible?  I have a book on vSphere, and based on what I've read, it looks like it is.

Yes - You can configure all ESX hosts into 1 cluster - though you'll need to manage access to storage (You won't want a VM hosted in site A using storage in Site B)   - where I have configured this scenario before, we used one site as production and one as DR (plus Dev / Staging) - VMs were approtioned to Datastores based on Failover policies as storage of course is only writeable on one side at a time . . .so we only want to access VMs on that storage form that site.

2. Will I need any other hardware beyond 4 Dell R510s, 2 EqualLogic PS4000s, and 4 good managed switches for the iSCSI traffic, with a dedicated fiber link between them, two large UPSs, and a few dozen CAT5 cables?  I already have a server to use for vCenter.

If you are hoping to migrate VMs between sites, you'll either need to have a strecthed VLAN (to allow the VM to run on the same IP address at both sites) - or need to consider a method of updating IPs for these VMs.

Previoiusly I have managed this by scripting failover from one site to another and setting all  IP addresses to DHCP  - but then created DHCP reservations for those mac addresses on both sides of the Network. the filaover script updated DNS records to make it work.

3. UPSs.  Is there a preferred brand that works well with vSphere?  I usually use APC units, but I don't have strong feelings about them one way or the other.

APC should be fine.

4. Which VMWare software will I need?  I'm a little hazy on what all the different packages do.  I'm guessing I need ESX Enterprise, and vSphere, but I want to make sure I'm not missing anything.

You'll need a vSphere license and ESX host licenses. Other than that - software  / licesning is that of an y monitoring / backup you'd like - and of course the normal licensing for the VMs. (OS etc)

5. What is the best backup method for this?  Is SAN replication sufficient?  Or, should I use an external tape unit?

I prefer Veeam  - but it depends on whether you need data duplication, or place in time backups.

Questions answered . . here is the extra bit I can contribute.

Think this through very carefully

Consider getting a consultant in for a day or 2 (you caould try talk you VMWare reseller into doing this for you)

Make sure you buy a version of ESX that has what you need . . not what you MAY need one day (no need to buy enterprise plus . . if you don;t need any of the add on features)

IP addressing accross sites will be your biggest challenge.

Document EVERYTHING - and make sure to keep your documents updated.

Good luck

[symen]

Yes - You can configure all ESX hosts into 1 cluster - though you'll  need to manage access to storage (You won't want a VM hosted in site A  using storage in Site B)   - where I have configured this scenario  before, we used one site as production and one as DR (plus Dev /  Staging) - VMs were approtioned to Datastores based on Failover policies  as storage of course is only writeable on one side at a time . . .so we  only want to access VMs on that storage form that site.

I hadn't thought of this, but it makes perfect sense.  Will this affect how vMotion works or anything?

If you are hoping to migrate VMs between sites, you'll either need to  have a strecthed VLAN (to allow the VM to run on the same IP address at  both sites) - or need to consider a method of updating IPs for these  VMs.

Happily, we already have stretched VLANs in place.

I prefer Veeam  - but it depends on whether you need data duplication, or place in time backups.

I saw Veeam mentioned a few times on here along with your post, so I took a look at it; it looks pretty nice.  Is backing up to a NAS device or similar with Veeam a good way to go?  Thinking more about it, I would like to have snapshots available in addition to the hot backup (I won't be managing every VM, so I'd like to be able to say, "Let me roll that back to the last snapshot for you" to the person who screwed up their VM to the point that it doesn't boot anymore).

Thanks very much for your help, your post was very informative.