Load Balancing or Clustering of Operating System in VMware products

[suri4u]

Hi,

I have a requirement which i need to plan it out :

Requirement is .. i need to build two physical host servers and install hypervisor for creating VM's on top of it.

We will create 4 VM's on each the physical hosts and those will be identical ( Eg. DC,file,print,utility servers on both the machines )

my requirement is :

1) Can we have load balance or clustering at the hardware level or Hypervisor level ,so that one server (eg., file server goes down ) other comes into picture ... just like a failover

Does ESXi support loadbalancing or clustering of Operating system.

Is there a way to do that in VMware ,if not ESXi any other paid product from VMware ?

Any suggestions ,inputs would be appreciated

Thanks

[abhilashhb]

Why not Go for OS Clustering, Like RHEL clusters and Windows clusters?

VMware has the feature of high availability that will help you bring up a VM on a different server when the underlying host goes down.You will also get application level failover methods from App vendors.

AFAIK there's nothing like clustering at ESXi Level. There's a feature called Fault tolerance which is supported only for VM's with 1 vCPU.

[suri4u]

My doubt is If we go for Windows Failover Clustering at the OS level , will the High Availability feature get into picture.

As we are going for failover clustering at OS level , high availability would do the same work as failover clustering .. right ?

Please correct me if i am wrong ?

Could you please explain me the windows clustering and high availability taking an example.

Thanks in advance .

[abhilashhb]

You have Server 1 and Server 2.(They are both in a HA/DRS cluster with shared storage)

Server 1 runs two VM's- VM1 and VM2 (They have Windows Clustering with some application running on them)

Server 2 runs two VM's- VM3 and VM4.

Now if Server 2 fails because of a hardware failure the HA will kick in and reboot the VM's 3 and 4 on the Server 1(This is possible because both the servers are seeing the same storage) - This is high Availability

Now If something went wrong with VM1 (Assuming it was the active node in the cluster) The cluster services failover to VM2 (Or VM2 takes over the ownership of services) - This is windows clustering

I would say use one of them. Do not use OS clustering and HA together.

What you could also do is make sure the VM's involving in Clustering are always kept on different hosts (VM1 on Server 1 and VM2 on Server 2) so that it will protect you against Host failures also.

High Availability protects you only when the Host itself goes down. It comes with some little downtime as the machines are restarted om other server(s). But Windows clustering does not involve downtime. As soon as the Node1 fails the Node2 takes over at the back end, the users willl not even know the difference.

Did that clear your doubt?

[john23]

HA clustering will give the application downtime. you can approach for Fault tolerance , but it has lot of constraints such as 1 vCPU support.

If you are going for windows VM, MSCS can help to achieve the application clustering.

Even for application monitoring you can check VMware heartbeat as well.

-A

[suri4u]

You have Server 1 and Server 2.(They are both in a HA/DRS cluster with shared storage)

Server 1 runs two VM's- VM1 and VM2 (They have Windows Clustering with some application running on them)

Server 2 runs two VM's- VM3 and VM4.

Now if Server 2 fails because of a hardware failure the HA will kick in and reboot the VM's 3 and 4 on the Server 1(This is possible because both the servers are seeing the same storage) - This is high Availability

Now If something went wrong with VM1 (Assuming it was the active node in the cluster) The cluster services failover to VM2 (Or VM2 takes over the ownership of services) - This is windows clustering

I would say use one of them. Do not use OS clustering and HA together.

What you could also do is make sure the VM's involving in Clustering are always kept on different hosts (VM1 on Server 1 and VM2 on Server 2) so that it will protect you against Host failures also.

High Availability protects you only when the Host itself goes down. It comes with some little downtime as the machines are restarted om other server(s). But Windows clustering does not involve downtime. As soon as the Node1 fails the Node2 takes over at the back end, the users willl not even know the difference.

Did that clear your doubt?

Thanks Abhilash for detailed explanation ..

We don't want to use shared storage beacuse if the shared storage goes down , both the VM's would be dummy. instead we want to go for local storage.

Exmaple : 2 physcial hosts with  192GB RAM,8TB HDD's

Here is the scenario, 2 Physical hosts running VMware hypervisor would have 4 VM's each with application like DC, File, Print and utility servers on windows server 2012. We would like to have High Availability or OS clustering within this VM's so that when one server goes down the other one has to come up (say if File server goes down on 1st host , file server on other host should come up with full functionality).

Can we implement this with VMWare ,if yes could you please provide me with some explanation and how to do it.

Could you also suggest me which product of vmware to go for ?

Does ESXi support HA feature or OS clustering ?

Please advice on how to proceed on this .

Thanks a lot in advance.

[tomtom901]

ESXi can provide HA (only with shared storage though) but in your scenario, I would configure OS based clustering / HA. For almost every Windows feature / role there is a native option for clustering. For the items mentioned in your example:

DC: Configure a second domain controller, but keep in mind the FSMO roles which can only be on one server at a time. If the domain controller holding the FSMO roles fails, and you can't recover from this faillure, you need to seize the roles from the server.

Fileserver: Create a fileserver cluster by creating a failover cluster in Windows.

DNS: Integrate the zones in Active Directory and use AD for the replication

DHCP (if applicable): Configure a split scope (WS 2012 has excellent features for this)

Print services: You can configure HA printing

You might need to look at providing some kind of redundancy for your custom applications, if you have any. Native Windows features or roles can be easily clustered or configured to provide HA support. Although you have a valid point on the SAN storage and when it fails, both hosts lose access to the storage and the VM's go down.

When you choose an enterprise grade SAN, which at least has redundant controllers and multiple disk group, a lot needs to happen before service interuption is introduced on the storage side. I would take a look at the roles and the services that you would like to provide on this group of server(s). Then check whether these services can be configured to provide HA support on an OS or feature base. If this is not possible for the important and required services, you might want to revisit the option for shared services.

Hope this helps,

Tom

[Paltelkalpesh]

If HA and FT is not suitable for your network then you have to do windows clustering on VMs, also check what storage using, because some of storage controller is not supporting to directly add LUN  as a RDM to VM.

go through  below link

VMware KB: Raw Device Mapping option is grayed out