ActiveDirectory and Exchange Server on ESXi

[kasper77]

Hi all,

for a project I am going to create a cluster of mail servers (let's call it ES ... and I need 2 mail server)

and a cluster of Web servers (let's call it WS and I need 2 web server).

Each cluster will have 2 ESXi host (VMware vSphere FT).

1) I have a question/doubt here about the cluster composition.

I was thinking something like this:

Cluster 1

- ESXi1: VM_ES1; VM_ES2;

- ESXi2: VM_ES1; VM_ES2;

Cluster 2

- ESXi1: VM_WS1; VM_WS2;

- ESXi2: VM_WS1; VM_WS2;

or should I have a solution like this (mixing the two type of Server)?

Cluster 1

- ESXi1: VM_ES1; VM_WS1;

- ESXi2: VM_ES1; VM_WS1;

Cluster 2

- ESXi1: VM_ES2; VM_WS2;

- ESXi2: VM_ES2; VM_WS2;

2) is it ok for have one VM for each server? or is common to combine more server on a single VM?

thanks a lot.

- -Kasper

[ThompsG]

Hi,

I would have one cluster with all 4 host in it and spread the VM's over those. Gives you better scheduling opportunities and if you have DRS then you can use Affinity/Anti-Affinity rules to provide high availability of services by keeping them apart so that if a host fails not all your Exchange/Web/whatever goes out at the same time.

Kind regards.

[fescaros]

Rather than using FT on Exchange you should be looking at DAGs, as they give you roughly the same amount of uptime SLAs with no performance hit.
Same in my opinion for Web Servers which should get their data from a backend and simply sit behind a load balancer, so that you spawn a new web server rather than trying to checkpoint the highly transactional CPU and network load on them (which is gonna cost you performance).

For obvious reasons licensing etc. might make FT more attractive but especially on a natural scale out workload like web servers it usually makes less sense than other alternatives.

[kasper77]

thanks a lot ThompsG!

so you suggest to have something like this:

Cluster 1

- ESXi1: VM_ES1;

- ESXi2: VM_ES2;

- ESXi3: VM_WS1;

- ESXi4: VM_WS2;

is this correct?

do I need some "spare" ESXi hosts for redundancy?

[BenLiebowitz]

You seem to be on your way for cluster planning but I wanted to throw this into the ring for your AD VMs.  I attended this session at VMworld and thought you might find some of the info handy.

VMworld 2016: VIRT7621 - Virtualize Active Directory, the Right Way! - YouTube

[ThompsG]

Hi there,

Yes something like that - I'm assuming here you have vCenter because you talk about VMware clusters. If that is that case then you make a HA enabled Cluster and if a host fails the VM running on it will be restarted on another host (assumes you have shared storage between all the hosts). That's why you separate the workload so that a host failure only impact one VM from a "service group" rather than the whole service. Once you have resolved the outage (planned or unplanned downtime) then you simple migrate the workloads back to gain full redundancy again.

Hope this makes sense. Also as others have mentioned I would not be using FT for Exchange/Web/Domain Controllers - there are other better solutions unless you absolutely must have 100% uptime but this is very really the case.

[kasper77]

Hi Matt and thanks for your answer!

I'm planning to have 4 ESXi host ...

2 for production (web server, mail server etc ...)

2 for management (vCenter server, etc.)

I will follow your advice and crete 2 cluster:

1 for the production (with HA, anti-affinity, etc ...)

1 for management

as well two network (2 physical switch too to completely separate the two network)

does it looks reasonable?

cheers

Kasper.