Failover clustering-IP address required

Question

Hi All,

I am bit confused ..Since in two node cluster  six ip address are required.

How many we need in case of four and five node cluster.Please explain??

Thanks in advance

---

Ritz

Answer 1

Depends on what you do... One IP address per physical node, one for the cluster itself and one more for each SQL instance is the minimum.

Answer 2

What abut Msdtc and private network???

Can you just list and give...

---

Ritz

Answer 3

You will need one public IP and one Private for each physical node. you will need one ip for the cluster instance and one for MSDTC.So, for a four node cluster, with four instances on them, would require 4*2= 8 for all physical nodes 4*1 =4 For all cluster instances 1*1 =1 msdtc ---- 13 IP Numbers

---

Hope it Helps!!

Answer 4

You need ips as for

1. 1 IP for Windows Cluster VIP
2. 1 IP for SQL cluster VIP
3. 1 IP for MSDTC
4. 1 IP for each machine as physical ips
5. 1 IP for  each machine as private\heart beat network

Refer Link : http://social.msdn.microsoft.com/Forums/en/sqldisasterrecovery/thread/19c144cb-d1c3-4789-bbe4-da899478384f

---

Regards,
Rohit Garg
(My Blog)
This posting is provided with no warranties and confers no rights.
Please remember to click Mark as Answer and Vote as Helpful on posts that help you. This can be beneficial to other community members reading the thread.

Answer 5

Practically, yes that is the minimum I would recommend, but...

2. You can choose to install (or not) one to many SQL FCIs -therefore one IP per FCI (in a non multi-subnet cluster)

3. MSDTC is not a requirement of a cluster OR even SQL Server OR even required to be clustered in SQL 2008 and upwards AND you can even configure one per FCI if required. However certain functions of SQL Server rely on MSDTC -such as distributed transactions. 

4. This is the public client access network of which you can choose to have several.

5. The private network is not a requirement of a cluster these days, especially if multiple "public" networks have been implemented (as per 4.) but it is still something I would tend to recommend.

---

Regards,
Mark Broadbent.

Contact me through (twitter|blog|SQLCloud)

---

Please click "Propose As Answer" if a post solves your problem
or "Vote As Helpful" if a post has been useful to you
Watch my sessions at the PASS Summit 2012

Answer 6

Re: private network. WSFC cares about having redundant networks to communicate over. Whether one is private and another public or both public, it doesn't reall care. Just clarifying the point.

So you do need 1 for the WSFC, 1 for each SQL Server FCI, 1 for each network on the local node. For a 6 node cluster:

1 - WSFC

1 - Single SQL instance (+1 for each additional)

6 - Network 1 on each node

6 - Network 2 on each node

Total: 14

If you are using iSCSI, that will be another 6 IPs (need a dedicated NIC and network for that on each node) for a total of 20, and if configuring DTC, one for each installation. For a single instance it'd be 1 IP address for  total of 21.

---

Allan Hirt Blog: http://www.sqlha.com/blog Coming in 2013: Mission Critical SQL Server 2012 - the followup to Pro SQL Server 2008 Failover Clustering

Answer 7

Re private network, agreed. Having redundant networks (and components that constitute them such as switches) as essential for a HA configuration, but thanks for elaborating on my reply. The point I was making was that in a scenario of only two networks being presented to a cluster, there is a strong argument that these should both be enabled and configured for client access to provide better availability for public network access rather than dedicating for intra cluster communications and hence being at risk of a single point of failure for client access.

---

Regards,
Mark Broadbent.

Contact me through (twitter|blog|SQLCloud)

---

Please click "Propose As Answer" if a post solves your problem
or "Vote As Helpful" if a post has been useful to you
Watch my sessions at the PASS Summit 2012

Answer 8

Hi there,

Let me count how many ip address involved in 2-node FCI first and then count for more.

1) You need a domain controller first. (count:1)

2) You need two nodes assume you have given physical address for each.(count: 2, total: 3)

3) You need a public network, and private network on each node(count: 4, total: 7)

You may have more networks on your nodes, but i am counting minimum requirement

4) You need an ip address for a cluster(count: 1, total: 8)

If you use any third party tools to connect shared disks or SAN i am not counting here to set up the environment.

5) Every node you add require 1-physical address, 2 network address a total of 3

These are minimum requirement to setup FCI.

Thanks

Kumar

Answer 9

If you don't count DCs, DNS, etc. (so only the cluster) my statement from above still stands. You can run DTC locally or within the SQL group, so techically no need for an IP on that. And you can run the cluster without a private network.

I am back to my question: What do you need? If you want to get to a fully fledged cluster you will need some more IPs. (Private network, DTC, Witness node or Fileshare Witness, etc.)

Answer 10

This is incorrect. For example, the node does not need a different address for public vs the one already assigned to it. My IP numbers above are correct based on the numbers given for a six node cluster. For a simple two node cluster it is:

1 - WSFC

1 - Single SQL instance (+1 for each additional)

2 - Network 1 on each node

2 - Network 2 on each node

Total: 6

---

Allan Hirt Blog: http://www.sqlha.com/blog Coming in 2013: Mission Critical SQL Server 2012 - the followup to Pro SQL Server 2008 Failover Clustering

Answer 11

@Allan: Still brings me down to 4 if you skip the second Network... Or did I miss anything there?

Answer 12

You need redundant networking. One network is a single point of failure - why bother clustering at that point? *Techincally* you could have a single network. I can also shoot myself in the foot, too. So you really can't skip it if you need to be highly available/mission critical.

And if you're using iSCSI (which wasn't indicated but let's throw it in), that would need its own network, too (that gets flagged in validation), so it could be as many as 8 IPs for a basic config.

---

Allan Hirt Blog: http://www.sqlha.com/blog Coming in 2013: Mission Critical SQL Server 2012 - the followup to Pro SQL Server 2008 Failover Clustering

Answer 13

Sure Thing. From a redundancy Point of view you are completely right. Although only from a high Level... Looking at actual implementations I have seen lately having more Networks doesn't even help anymore in most cases... but this has nothing to do with this thread. Maybe we find some time next week to Chat about it in Detail.

Rick

Answer 14

Also consider the case of WSFC made up of VMs, which are unlikely to see any availability benefit from configuring additional virtual network adapters, and that Windows 2012 now natively supports NIC Teaming so that a single host Network Adapter can span multiple adapters and Ethernet switches.

David

---

David http://blogs.msdn.com/b/dbrowne/

Answer 15

WSFC of VMs really only has three scenarios under any hypervisor:

1. All nodes under the same hypervisor, implement to account for patching the OS.

2. AG deployment in SQL Server 2012 with standalone instances since no shared storage (much easier for virtualization)

3. Doing a "proper" WSFC/FCI but splitting nodes across multiple hypervisors to ensure no single point fo failure.

You need to think of the pros and cons here ... WSFC under a hypervisor isn't neccessarily a slam dunk.

---

Allan Hirt Blog: http://www.sqlha.com/blog Coming in 2013: Mission Critical SQL Server 2012 - the followup to Pro SQL Server 2008 Failover Clustering

Answer 16

More than happy to talk about it next week. You know where I'll be :)

As an aside, this is why it's a lot of "it depends" with all of these architectures. There's always more to the story.

---

Allan Hirt Blog: http://www.sqlha.com/blog Coming in 2013: Mission Critical SQL Server 2012 - the followup to Pro SQL Server 2008 Failover Clustering

Answer 17

Couldn't agree more Allan...